01 · Two steps
Traffic didn't grow steadily. It jumped twice: in early August and on Oct 21.
8K → 110KDaily web visits, just before and after Oct 21
VerifiedThird-party estimate · Details →
The short version: six findings
Teamily's growth came in two paid steps, reached four very different audiences, and is described with numbers that visible data doesn't support. Each card links to the section with the evidence.
02 · What drove them
The first step was bought with Google search ads. The second lined up with paid creators.
$280KEst. Google search spend in August, up from $3K in July
InferredTimeline → · Channels →
03 · Who the users are
Not one ICP: US desktop, Taiwan, South Asia and Africa, and Europe each came in a different way.
86%Of web visits from the US in the last 28 days. Taiwan rose from 0.5% to 3.3%.
InferredUsers → · Countries →
04 · Claimed vs visible
The headline numbers are far above what can be seen in public data.
250M vs 14MClaimed monthly social views vs publicly visible October views on X, Threads/Facebook, YouTube and TikTok (~6%)
05 · Marketing vs usage
Marketing sells collaboration. Users come for outputs: research, slides, video and websites.
61% vs 14%Share of feature content on collaboration vs on the top 4 used features
VerifiedBy tagging · Features →
06 · Lessons
Copy the tactics: search ads as a demand test, one message per market, one keyword per creator. Not the numbers.
100Peak weekly search for “teamily ai” (Oct 25), with a paid creator, not a press release
InferredLessons →
01 · Summary
Not a viral breakout: web traffic rose in two steps, first from bought Google search ads, then from paid creators plus US desktop traffic with no clear source.
Summary
This wasn't a viral breakout. Web traffic rose in two steps. The first was mostly bought Google search ads. The second came from a batch of paid creators plus US desktop traffic with no traceable source. Most of the big public numbers are inherited from an older product or come from paid exposure.
Cause and effect, in order
Feb–JulBarely any traffic7K–42K visits a month. PR and wire releases drove no search Verified
→
Aug · Step 1Bought search + likely junk trafficEstimated $280K on Google search. Visits up 11× to 470K. Italy: 11% of visits, near-zero time on site Inferred
→
Mid-Aug–SepYouTube review waveEnglish "Teamily AI Review" videos: 417K views in September, 21 with invite codes Verified
→
From Oct 21 · Step 28K → 90K–110K visits a dayTaiwan creator Bagel, India "ai agent" ads, and small IG accounts all ran at once, yet 86% of traffic is still US Inferred
→
Nov 3Pre-fundraise victory lap32× claim, three paid wire releases at once, templated reposts from low-follower X accounts, No. 2 on Product Hunt Verified
How Teamily grew: 6 findings
- The first step was bought. Visits went from 42K in July to 471K in August, almost all on desktop. That month SimilarWeb estimates $280.5K in Google search spend (July: $3K), on competitor names (perplexity, timely ai) and broad terms like "free AI video/image tool." Verified (spend is a third-party estimate). From Aug 15, daily visits fell back to about 2K. Italy made up 11% of August visits with a 99.5% bounce rate, so part of it looks like bot traffic Inferred.
- September held up on YouTube reviews and continued ad spend. About 38 English reviews with ~780K visible views, mostly mid-August to late September. Google spend was still $147.6K in September. Verified
- The second step (from Oct 21) has a few known drivers, but they don't explain all of it. In the same window: paid Taiwan creator Bagel (23 posts, ~1.53M views; "teamily ai" search peaked for the year in the week of Oct 25), India "ai agent" search ads (from Oct 26), and templated posts from small IG/TikTok accounts. Yet the US still accounts for 85.9% of visits in the last 28 days (1.16M), and Taiwan only 44.5K. Most of the step has no identified source Inferred.
- Users fall into four groups by market. US desktop (bought search traffic plus logged-in daily use; 80%+ on web). Taiwan (hooks like "free websites," "side hustle," "quit your job"; low intent; No. 2 in the last 28 days). South Asia and Africa (people searching for free tools; India has the most ads of any country). Europe (creator-run ads, small volume). The official "<20-person small business" segment makes up only ~3% of 1,165 external posts Inferred (from tagging).
- High direct traffic (93% in October) has a simple explanation. The product itself runs on teamily.ai (subdomains are 99.89%), so every visit by a logged-in user counts as direct. In September, visitors viewed 3.5 pages and stayed 3:39 on average, which looks like product use. Links in creator DMs and in-app browsers that drop the referrer also land in direct Inferred.
- What to discount: ① User counts keep shifting (2M → 3M → 4.5M → 5M+, all from ChainOpera or the "product stack"; Teamily's own numbers go 10K → 320K). ② Trending on X came from paid quote posts (14 of 20 captured quotes on Oct 8 are labeled Paid) plus templated reposts from low-follower accounts on Nov 3. ③ "250M social views in a month" and "150K people commenting for the link in a month" are tens of times to ~7× above visible data. ④ PR is almost all paid wire releases; it serves fundraising and credibility, not user acquisition.
Traffic steps and drivers
Bars = SimilarWeb monthly visits (third-party estimate; October = sum of daily data, as the monthly report isn't out). Orange line = SimilarWeb's estimated Google search ad spend, available only for Jul–Sep. Labels = drivers visible in the same period; causation is inferred.
Daily visits (Jul 18–Nov 1). Shaded bands = drivers in the same period. The steps switch on and off abruptly, and each step is about 90% direct, which looks more like ad batches starting and stopping than organic word of mouth Inferred. The Oct 11–20 dip is unexplained.
Key metrics
Company-reported (unverified) Third-party estimate / publicly visible
10K → 320KSign-ups in "one quarter," 32× (reported Nov 3; the quarter's dates don't match other reported numbers)
20K / thousandsDAU / paying users ("several thousand," ~1–2% of sign-ups)
250M+Social views in a month (visible: YouTube ~780K, TikTok ~27K)
150K+Comments asking for the link in a month (visible IG comments total 21,360)
1.43MWeb visits in October (SimilarWeb; September = 14.9× July)
$280KGoogle search spend in August (SimilarWeb estimate; ~$430K over 6 months)
3.29%Taiwan's share of visits, last 28 days (44.5K, No. 2; only 0.16% in September)
≈1/10Brand search interest is about 1/10 of Genspark's (Trends "teamily ai")
Inherited user counts, for comparison
Log scale; all numbers company-reported. Gray line = two figures in the same Jul 8 Business Wire release differ by about 250×. The big numbers come from ChainOpera's crypto user pool Inferred (strong evidence: chainopera.ai gets only 1.5K–3K web visits a month).
02 · Actions
Search for “teamily ai” peaked in the week of Oct 25, alongside a paid Taiwan creator; no press release or the Nov 3 letter moved search.
Action timeline: what they did → how traffic moved
Search for "teamily ai" peaked for the year in the week of Oct 25, alongside Taiwan creator Bagel's paid posts Verified. The "teamily" spike in the week of Oct 4 came from K-pop group &TEAM's mascot of the same name, not Teamily. Web traffic rose in two steps: August (about $280K in Google search ads) and from Oct 21 Causation inferred. Neither the press releases nor the Nov 3 shareholder letter produced a search peak.
Top: Google Trends weekly index for "teamily ai" (the single word "teamily" is polluted by K-pop group &TEAM's mascot of the same name, so it's not used; each region's line is normalized on its own, so compare shapes, not heights; values are relative 0–100, 0 = below the reporting threshold; no weekly data before Nov 2025). Bottom: one lane per row. Large dot = major action, small dot = minor action, hollow = inferred, bar = campaign span, curved arrow = link between actions (toggle in the legend). Click any action for details; click a point on a traffic line to see what happened 7 days either side. Drag the slider at the bottom to zoom. All company numbers are self-reported.
Major actions
| Date | Lane | Action | Summary | Confidence |
|---|
03 · Timeline
For 13 months, new product shells, audiences and wire releases brought no real traffic; the volume came in two steps, in August and from Oct 21.
Growth timeline
For the first 13 months, the team kept changing the product shell and the audience and sent out wire releases, with no real traffic to show for it. The real volume came in two steps: August (bought search) and from Oct 21 (paid creators plus US traffic from unknown sources).
Each dot is a dated event; hover for details and sources. Hollow dots are traffic or search data (SimilarWeb, Trends "teamily ai"). October is crowded, so overlapping labels hide automatically; drag the slider below to zoom in. User counts are company-reported; visits are third-party estimates.
Five stages
| Stage | Dates | What they actually did | Stated target audience | Visible results |
|---|---|---|---|---|
| 0 Predecessor: ChainOpera | Jun – Sep 2025 | Multi-agent group chat; sponsored post in The Block; COAI listed on Binance Alpha, points airdrop | Crypto users | Claimed "2M+ sign-ups"; chainopera.ai gets only 1.5K–3K web visits a month Inferred: points/airdrop accounts |
| 1 Cold start | Feb – Apr 2026 | Invite codes; Chinese media coverage on the same day (Feb 13) + paid English wire release; launch wave from X AI-tool accounts | Broad audience, "a WeChat for AI" | 10K+ beta users (company-reported); only a tiny signal in Trends Verified |
| 2 Going global | May – Jul 7, 2026 | iOS/Android launch; Singapore wire release; broad Google ads begin (May 22) | Singapore professionals and creators → startup teams | 10K–40K visits a month; almost nothing from Singapore; zero search for "teamily ai" in May–July Verified |
| 3 Scale-up | Jul 8 – Oct 31, 2026 | In July, official accounts go quiet and spend shifts to Google search (estimated $280K in August); YouTube reviews mid-Aug to Sep; in October, Taiwan creator Bagel, templated posts from small IG/TikTok accounts, creator-run ads in Europe, India "ai agent" ads; offline buses and trade shows | Small businesses (<20 staff) / solo businesses (official); actual messaging varies by market: side hustles and free websites in Taiwan, making money on the side on IG | Monthly visits 42K → 470K (Aug) → 620K → 1.43M (Oct); "teamily ai" peaks in the week of Oct 25 Verified |
| 4 Victory lap | Nov 1, 2026 – | Shareholder letter + 32× long post; three paid wire releases; templated reposts from low-follower X accounts; ~70–114 posts from small #teamilyaipartner IG accounts in 3 days; first Meta brand ads; No. 2 on Product Hunt | Investors + the above | Company-reported 320K sign-ups, 20K DAU, thousands of paying users; no global search spike Verified |
There was no "biggest breakout of the year on Oct 6." That day K-pop group &TEAM launched a mascot also named "TEAMILY," and fan searches polluted the single word "teamily." The clean term "teamily ai" took its first step Oct 4–17 (57, 61) and peaked for the year in the week of Oct 25 (100), matching Taiwan creator Bagel's Facebook post Verified.
X posts: lots of views, few likes
How to read this Inferred
The two posts with both view and like data got only about 3 and 13 likes per 10K views.
Views and engagement don't match, which looks like paid promotion. Supporting evidence: of the quotes captured on the Oct 8 and Oct 28 posts, 14 of 20 and 33 of 40 carry the X "Paid partnership" label Verified. Much of the claimed "250M+ social views in a month" may be this kind of paid exposure Inferred.
04 · Playbook
One loop: creator content finds the hook, paid scales it, and conversion data shapes the next brief; markets start wide and narrow by paid conversion.
Playbook and process
One loop: creator content finds the hook, paid scales it, and conversion data feeds the next brief. Markets and channels start wide, then narrow by paid conversion.
Source: Summer Chen's long post "32× Growth in One Quarter With a Small Team" (Nov 3, 2026). All numbers below are company-reported.
1. The core growth flywheel
Main loop Verified (original: "We ran ours as one loop"). Videos that beat the threshold are flagged by the UGC Agent for the paid team to scale. The gray branch is the attribution gap Summer herself admits.
Paid platforms: Google search ads are the largest measurable paid source (SimilarWeb estimates $280K in Aug, $148K in Sep). They target competitor and free-tool keywords, not hooks found by creators Verified. So the "creator hook → paid scale" loop doesn't match what runs on Google Inferred. Meta brand ads only started Nov 3; paid social is near zero in SimilarWeb; the ChatGPT pixel is installed, but it's unknown whether ads are running.
Second loop (in-product referral): two-sided invite credits / free Plus, affiliate commissions, "Sponsor your group's AI usage", and forking other people's agents. These modules exist in the front-end code Verified; there's no data on how many sign-ups they drive.
2. How markets were chosen
Step 1Test 20+ markets at onceEach market tests paid + creators + UGC, including North America, Taiwan, and Latin America
→
Step 2Rank by sign-upsSummer calls this "the obvious mistake": sign-ups only show curiosity
→
Step 3Re-rank by paid conversion"The map changed": payment shows the product solved a real problem
→
Step 4 (about 1 month in)Focus budget on the top 5–10"Test wide. Measure fast. Fund narrow."
→
OngoingReview and reallocate weeklyFunnel: views → CPM → visits → sign-ups → activation → paid; budgets stay flexible
Process Verified. The post doesn't say which markets were cut at each step or give per-market conversion rates, so no counts are shown here.
3. Nine pilot channels in month one
Core (results described) Verified (claim)
Creator partnerships Screen 100, keep 5 (partly matches); Taiwan was really one paid creator, Bagel
UGC Claims 150K+ link-request comments in 1 month; visible IG comments total 21,360
Paid ads Google Search: est. $280K in Aug, still running in Nov
Offline / events (with examples) Verified
Offline Small early budget; 130 buses for SF Tech Week
Events About 200 attendees → 5M online reach (company-reported)
Rule: if the audience isn't ICP, the event is unrelated to the product, or the goal is unclear → cut it, "even if the attendance number looks impressive".
Listed only (no results data)
SEO Mostly brand terms; sitemap has only 10 URLs
GEO A focus: llms.txt + AI crawlers allowed
Campus No data
Referral In-product invite/commission system; no contribution data
Summer only lists the names. The notes on the right come from site code, robots.txt, and llms.txt Verified; "SEO is brand terms only, GEO is the focus" is a judgment Inferred.
4. Case study and cadence
Taiwan: a low-cost market reached #3 Company-reported, partly matches
<$20KTotal spend
50+ → #3Market rank (by users)
"With less than $20K in total spend, it went from outside our top 50 markets to our #3 region by users."Summer Chen, X, Nov 3, 2026
"Creator CPMs were lower, audiences had less AI-content fatigue, and strong product stories had more room to travel."Summer Chen, X, Nov 3, 2026
Note: the rank is by users, not paid customers. External data: Taiwan's share of web visits went from 0.16% in Sep to 3.29% in the last 28 days (#2, SimilarWeb). Taiwan's search peak was driven mostly by one paid creator, Bagel (23 posts on Threads/FB/IG, about 1.53M views, #ad). The hook was free websites and side hustles, not small business Verified.
Execution cadence Verified
100 → 5Creators screened
2–3/weekFeature launches
20+ × 7+Markets × channels
- Creators: screen, negotiate, review scripts, check product accuracy, repeat across markets.
- Content: a workflow turns product updates and existing assets straight into launch videos.
- Reporting: a 7-day funnel report is pushed to the group automatically; the UGC Agent posts daily creator count, spend, views, and CPM.
- People: a small team, no large growth org Team size not disclosed
5. In Summer's words
"Sign-ups showed us where curiosity was. Paid conversion showed us where the product was solving a problem worth paying for."Summer Chen, X, Nov 3, 2026
"Creator content discovered demand. Paid amplified it. Conversion data improved the next round of content."Summer Chen, X, Nov 3, 2026
"Search or paid search might receive the last-click credit, even when creators and UGC created the original demand."Summer Chen, X, Nov 3, 2026
"The 200 were the event. The 5M were why we did it."Summer Chen, X, Nov 3, 2026
05 · Users and ICP
Only search ads, YouTube reviews and comment-to-get-link funnels show measurable volume, and users split into four groups by market, not one ICP.
User sources and ICP
Only three sources show measurable volume: Google search ads, YouTube reviews, and the creator comment-to-get-link DM funnel. Paid X creators buy reach, PR buys credibility, and the million-plus user counts are inherited. Users split into four groups by market, not one ICP.
User and team counts are company-reported. Visits, country shares, and ad spend are third-party estimates from SimilarWeb / Semrush. Views and comments are publicly visible platform counts. The company hasn't published sign-ups by channel.
1. Where users come from
Line width = a subjective score (1–5) of evidence strength × estimated contribution, not real traffic. Green = visible data; orange = inferred; gray = destinations that produce no sign-ups. The middle column shows which SimilarWeb channel the traffic lands in. Creator links mostly open in in-app browsers or DMs, so the source is lost and recorded as Direct. The product itself runs on teamily.ai, so logged-in usage also counts as Direct.
2. Sources ranked by evidence
| # | Source | Evidence | Who it reaches | Assessment |
|---|---|---|---|---|
| 1 | Google search ads | SimilarWeb spend estimate: $3.0K in Jul → $280.5K in Aug → $147.6K in Sep; 89K paid search visits over 6 months, about $4.84/visit; 137 ads across 152 countries; keywords = competitor names + free AI tool terms; #1 for "ai agent" in India since Oct 26 | US desktop, India / South Asia | Verified heavy spend; the main cause of the Aug step Inferred |
| 2 | English YouTube reviews | About 38 videos, about 780K visible views (417K in Sep); 21 carry an /i/XXXX invite code, 9 are labeled paid promotion | English-speaking AI tool fans (South Asia, English-speaking Africa) | Largest creator channel by views Verified; sign-ups unknown |
| 3 | Creator comment keyword → DM link | Visible IG comments total 21,360: gptsavyy 7,738, kats.workspace 5,378, kiwi4work 1,890, jerry.juuu 1,789; keywords TEAMILY / team / 網站 (en: website)…; askgpts uses ManyChat to send links automatically | Young people, side-hustle seekers | Tactic verified; clicks land in Direct Inferred |
| 4 | Taiwan paid creator Bagel | 23 posts on Threads/FB/IG, about 1.53M views and 5,200 comments, all #ad; FB posts on Oct 24–28 match the year's peak for "teamily ai"; Taiwan had 44.5K visits in the last 28 days (#2) | Taiwan office workers, side hustlers, people looking for free websites | Main driver in Taiwan Verified; only about 3% of total visits |
| 5 | IG partner micro-account network | 114 #teamilyaipartner posts from 20 accounts (most <1K followers), 5–11 posts each on Nov 2–4; tagged with US and Canadian cities | Young people in the US and Canada (side hustles, selling sites, coloring books) | Pay-per-post broad reach, late and small Inferred |
| 6 | Creator-run Meta ads (Europe) | 5 ads, 1–5 days each: Collin in 26 EU countries (EU reach 16,157), Antonio in Italy, mehdi in France | Marketing/ad agencies, content creators | Small Verified |
| 7 | In-product /i/ invite codes | Two-sided credits, free Plus for invites, affiliate program; all creator links carry invite codes; Telegram has credit-farming posts | — | Amplifier + source of low-quality sign-ups; no contribution data Inferred |
| — | Paid X creators / engagement pods | 77 of 604 posting accounts labeled Paid; the 198 quoters of two Nov 3 posts had median follower counts of 14–24; X ≈ 0 in SimilarWeb social referrals | — | Buys reach, not users Inferred |
| — | PR wire releases | 37 of 60 overseas articles are wire releases; no search spike after the Jul 8 or Nov 3 releases; most backlinks die after 2–3 months | Investors, AI search training data | Not an acquisition channel Inferred |
| — | Inherited from ChainOpera | 2M → 5M+ user claims; chainopera.ai gets 1.5K–3K web visits a month | Crypto points users | Only props up the big public number Inferred (strong evidence) |
No evidence found for: Hacker News, Reddit (all results are &TEAM), the Discord community (467 members), Apple Search Ads, paid social (68 visits in 6 months per SimilarWeb).
3. ICP: four groups by market
US · desktop
| Volume | 79.6% of web visits Apr–Sep; 85.9% in the last 28 days (1.16M visits) |
| Source | Google competitor and free-tool keywords; official X, offline; IG micro-accounts only started in Nov |
| Message | Small businesses with <20 staff, AI teammate, solo business |
| Caveat | Mostly desktop with abrupt on/off steps; automated traffic can't be ruled out Inferred |
Taiwan
| Volume | 0.16% in Sep → 3.29% in the last 28 days (#2); highest search share worldwide; #84 in iOS Productivity |
| Source | Bagel's paid posts + micro-accounts like kiwi4work and gptsavyy; not search ads |
| Message | Free website lists, "don't want to go to work," side hustles, "an AI LINE group"; not one post about small-business teams |
| Caveat | Mostly low-intent sign-ups Inferred; Summer's "<$20K to #3" only partly matches |
South Asia / Africa
| Volume | India 6.0% Apr–Sep; Pakistan 4.4% in Sep; Bangladesh 2.0% in the last 28 days; #82 on iOS in Nigeria |
| Source | Most Google ads (IN 46, PK 27, NG 26); English YouTube reviews |
| Message | "AI PPT/Doc Generator — Free", "ai agent", internship report generator |
| Caveat | People searching for free tools; low willingness to pay Inferred |
Europe
| Volume | <1% per country; Italy's 11% in Aug was zero-dwell traffic |
| Source | Serial AI-tool promoters running their own Meta ads; Google targeting FR 28, DE 26 |
| Message | Ad agencies using 3 agents to analyze client ads; French "make money with a system" |
| Caveat | The creators' audience, not a precise ICP Inferred |
Content tagging (1,165 posts, inferred): 51% of Instagram content targets "students / side hustles"; X mostly targets "AI early adopters" and "workplace teams"; "small-business owners" are only about 3% of all content (37/1,165). Summer says Latin America was tested, but web traffic shows only Brazil at 0.5%.
4. How the public positioning changed (official messaging)
2025Crypto users
"Agent Social Network"Points/airdrops to attract users
Feb 2026Everyone
"WeChat for AI"Family / friends / community / work
May 2026Singapore professionals / creators
"Singapore's professionals and creators"Almost no spend afterward
Jul 2026Startup teams
"building and growing a company"From idea to GTM and fundraising
Oct 2026Small business / solo business
"98% have fewer than 20 employees"Creators barely followed this message
Each step has a source quote Verified. But this is public positioning; the acquisition message differs by market (see above).
5. Product and pricing
Pricing (website) Verified
| Free | Up to 5 agents, limited messages; watch ads for more |
| Plus | $19.99/month (students $9.99) |
| Premium | $99.99/month |
| Pro | $199.99/month, includes AI video |
| Other | Annual plans get 2 months free; in-app prices are higher (Plus $28.99, Pro $289); Contact Sales |
The 19.9 / 199.9 in the Feb Chinese-language article is old pricing. Paid users are "several thousand," about 1–2% of sign-ups Inferred, based on company-reported figures. No named customer case studies and no customer logos on the homepage.
What users actually use
Top 4 uses in the shareholder letter: research, PPT/docs, video, and website building. Yet 63% of marketing content talks about "group chat / AI twins / multi-agent / memory," and only 13% covers these 4 output features Verified (by tagging). Google ads also only sell tool keywords (AI PPT / Doc / Website Generator), not "group chat."
ResearchPPT / docsAI videoWebsite builder (teamily.run)Daily briefing
Two-sided /i/ invite codesAffiliate commissionsPay for a group's AI usageCheck-ins / leaderboards
SimilarWeb audience: 70.5% male; 27.8% aged 25–34, 23.7% aged 18–24 (6 months).
6. Where Google ads are targeted
137 Google search text ads, counted by the regions listed on each ad's detail page (one ad can target several countries). These are ad counts, not spend. India has the most ads, but almost nobody in India searches the brand → it's buying broad-keyword sign-ups Inferred.
06 · Countries
Search and app ranks point to Taiwan, South Asia and Nigeria, but over 80% of web traffic is from the US; the two data sets measure different users.
Countries and regions: where and when
Search and app ranks point to Taiwan, South Asia and Nigeria, but over 80% of web traffic (SimilarWeb) is from the US. The two data sets measure different users.
Taiwan 100#1 in “teamily ai” search share (Hong Kong 73, China 66, Nigeria 24, US 3)
Jul → OctTargeted Google ads: North America + South Asia + Africa → Germany/France + Indonesia/Philippines → many EU countries / LatAm / Taiwan, HK, Japan, Korea
US 80% → 86%SimilarWeb share of web visits (Apr–Sep → Oct); Taiwan went from 975 to 45K visits, Sep → Oct
iOS #82 / #84Nigeria / Taiwan free Productivity chart; US Android only #1,137
All numbers are third-party estimates or company-reported. Search uses only the clean term “teamily ai” (the bare word teamily is polluted by the mascot of the K-pop group &TEAM).
① Market × month: one signal at a time
Data not ready
● Dot = a company claim or news action in that market and month (hover to read).
② Weekly “teamily ai” search by country (each scaled on its own; read shape and start only)
Vertical lines: May 26 Singapore · Jul 8 official launch · Oct 8 2.0 opens · Nov 3 shareholder letter / Product Hunt. Mexico, Japan and Korea are 0 all year (not shown).
③ Search share vs web traffic share: two different maps
Data not ready
Left: Google Trends regional share (12 months, low-volume countries included; based on share of searches within each region, so small regions are inflated). Right: SimilarWeb share of visits; gray = Apr–Sep combined (Aug–Sep are 93% of visits), purple = Oct 5 – Nov 1.
④ Markets at a glance
| Market | First signal | Peak | Channels seen | Language | Main pitch / audience | Evidence | Notes |
|---|---|---|---|---|---|---|---|
| TW | 2026-02 | 2026-10 | Search/Web traffic/Google ads/Creator content/Creator views/PR/App rank | zh-Hant | “AI version of a LINE group”; side hustle / office workers / owners cutting costs; comment “AI” or “team” to get the link | Strong | Bagel and others: 25 Threads/FB #ad posts in October (top post 890K views); Google ads only targeted Taiwan in October (4); SimilarWeb 0.47% for Apr–Sep, 3.3% in October |
| HK | 2026-07 | 2026-10 | Search/Google ads/PR/App rank | zh-Hant/en | Follows Chinese wire releases (Business Wire zh-HK) + a few reviews | Strong | One Google advertiser is TensorOpera HK Limited (21 ads); this does not mean ads ran in Hong Kong |
| CN | 2026-02 | 2026-02 | Google ads/Creator content/PR | zh-Hans | “AI WeChat” / group chat for people + agents; story for investors, talent and the Chinese diaspora | Medium | No direct payment in mainland China (no WeChat Pay/Alipay); Trends CN share of 66 likely comes from overseas Chinese / VPN users [Inferred] |
| US | 2026-02 | 2026-11 | Search/Web traffic/Google ads/Creator content/Creator views/PR/App rank | en | Small businesses under 20 staff, solo businesses, startup teams; AI teammate | Strong | Google targeting US/CA from July; official home market + offline events; SimilarWeb 80–92%, but clean-term search share only 3 |
| CA | 2026-07 | 2026-11 | Web traffic/Google ads/Creator content/Creator views | en | Same as North America | Medium | 7 Google ads; North America is a market Summer named |
| EU-DE | 2026-07 | 2026-11 | Web traffic/Google ads/PR | en/de | Google search ads (no German creators) | Medium | Google targeting DE/FR from August, widened to many EU countries in October; no German creators |
| EU-FR | 2026-02 | 2026-10 | Google ads/Meta ads/Creator content/Creator views/PR | fr | French creator “Commente GAGNANT” (system to earn money / side hustle) | Strong | Google targeting DE/FR from August; one French creator ran his own ads on Oct 29 |
| EU-other | 2026-02 | 2026-10 | Web traffic/Google ads/Meta ads/Creator content/Creator views/PR/App rank | en/it/es/pt | Ad agency uses 3 agents to analyze client ads (Collin, 26 EU countries) | Strong | October Google targeting IT/ES/NL/Nordics (20 ads); Italy's 52K visits all in August with 99.6% bounce rate, while Google did not target Italy in August → likely bot traffic [Inferred] |
| GB | 2026-02 | 2026-11 | Web traffic/Google ads/Creator content/Creator views | en | English AI tool accounts | Medium | 4 Google ads; creators mostly mixed UK/US |
| IN | 2026-02 | 2026-11 | Search/Web traffic/Google ads/Creator content/Creator views | en | Broad search ads for AI agent / free AI video, slides, report generators; Indian English AI tool accounts | Strong | Google targeting every month Jul–Oct (9–10 ads/month), the steadiest; Semrush IN paid keyword “ai agent”; search first appears in the week of Oct 11 at its peak |
| PK | 2026-07 | 2026-11 | Search/Web traffic/Google ads | en/ur | South Asian review accounts + broad-term traffic (inferred) | Medium | Google Jul–Sep often bundled with IN; SimilarWeb 2.7%; Trends appears in late October |
| ID | 2026-05 | 2026-11 | Search/Web traffic/Google ads/Creator content/Creator views | en/id | Google search ads | Strong | Google targeting from August (often bundled with PH); Trends only the week of Oct 25 |
| PH | 2026-08 | 2026-11 | Search/Web traffic/Google ads | en | Google search ads | Medium | Google targeted ID+PH in August; Trends late Oct – Nov |
| SEA-other | 2026-05 | 2026-05 | Web traffic/Google ads/Creator content/Creator views/PR/App rank | en | Singapore professionals & creators (May PR) | Strong | Almost no trace after the May PR: one Google ad targeting SG/MY in October, search ≈0; Android Thailand #287, Malaysia #580 |
| LatAm | 2026-07 | 2026-11 | Search/Web traffic/Google ads/Creator content/Creator views/PR/App rank | es/pt | Summer named it as tested | Strong | Google targeting only from October (AR/BR/CO/Spanish-speaking); Android Mexico #717; almost no creator content |
| Africa | 2026-02 | 2026-09 | Search/Web traffic/Google ads/Creator content/Creator views/App rank | en/ar | Arabic X invite posts in February; X reply accounts in November | Strong | Google Jul–Sep targeting EG/KE/NG; iOS Nigeria #82, Kenya #148; the NG one-week search spike on Sep 20 is noise |
| JP | 2026-02 | 2026-05 | Web traffic/Google ads/Creator content/Creator views | ja | — | Medium | The bare word teamily is polluted by &TEAM; clean term 0 all year |
| KR | 2026-10 | 2026-10 | Google ads | ko | — | Weak | &TEAM's home market; clean term only one noisy week in Nov 2025; 2 Google ads in October; bought Korean wire backlinks in July (for SEO, not users) [Inferred] |
| other | 2026-02 | 2026-11 | Web traffic/Google ads/Meta ads/Creator content/Creator views | en | English AI tool accounts with no location / Worldwide ads | Strong | AU/AE/BD etc. + unlocatable |
⑤ Conclusions
- Sequence: February Chinese PR + English X invite posts (North America story) → May Singapore got one wire release and left almost no search or ad trace → July targeted Google ads start (US/CA + India/Pakistan + Egypt/Kenya/Nigeria) → August adds Germany/France and Indonesia/Philippines, and web traffic jumps in Aug–Sep → October ads reach many EU countries, LatAm, Taiwan, HK, Japan and Korea; Taiwan leads search and India first appears in the week of Oct 11 → late October the Philippines, Pakistan, Indonesia and Nigeria follow in search. Verified
- Taiwan has the strongest search + app signals: clean-term share 100, iOS Productivity #84. The driver is Threads/FB creator #ad posts (one Bagel/bagelcuts post on Oct 5 had 890K views: “AI version of a LINE group”, “don't want to go to work”, swap links in the comments) plus templated IG/Threads small accounts. This matches Summer's “under $20K to reach #3”. Verified; the spend figure is company-reported.
- From July the steadiest Google search ads ran in India (9–10 a month); Taiwan got only 4, and only in October. Taiwan's volume was not bought with search ads. Verified Broad South Asian terms (ai agent, free AI video/slides) may bring many low-intent sign-ups. Inferred
- SimilarWeb's “US 80–86%” conflicts with search and app data. It is mostly desktop web (82% desktop for Apr–Sep), so it misses Taiwan and South Asia users on phones and in the IG in-app browser. US desktop traffic jumped 33× from July to August, while SimilarWeb estimates August Google search spend at $280K (competitor and free-tool keywords), plus likely junk traffic. Inferred It does catch part of the shift: India was 64% in July (the month Google targeting started), and Taiwan rose to 3.3% in October. Verified (third-party estimate)
- Europe: Google tested Germany/France first in August, then reached many EU countries in October, plus creators running their own ads (Collin in 26 EU countries leaning to low-CPM ones, French creator mehdi). Volume is small. Verified Italy had 52K web visits in August with a 99.6% bounce rate, which looks like bot traffic. Inferred
Data status: trends_clean_weekly.csv: ok; sw_countries_monthly.csv: ok; google_ads_by_country.csv: ok; creatives.csv: ok; meta_creator_ads.csv: ok; creator_content.csv: ok; other_social_posts.csv: ok; pr_coverage_cn.csv: ok; pr_coverage_global.csv: ok. Rebuild: python3 -I data/geo/build_geo.py --lang en.
07 · Channels
Most of the money goes to creators/UGC and Google search ads; official accounts are small, and reach comes from creator distribution.
Channel by channel: where the money goes
Most of the money goes to creators/UGC and Google search ads. The official accounts are small; reach comes from creator distribution.
Sources: website source code, Google Ads Transparency Center, Meta Ad Library, SimilarWeb, app stores (collected Nov 2026). Company numbers are company-reported; SimilarWeb figures are third-party estimates.
1. Channel evidence matrix
Verified
Inferred / indirect data only
No evidence
Importance = this report's judgment Inferred
| Channel | Evidence of spend | Results data | Importance |
|---|---|---|---|
| Creators / UGC | #ad, Paid partnership; screen 100 to keep 5, negotiate rates, review scripts | One UGC campaign drew 150K+ link requests in a month; Taiwan reached #3 on <$20K (company-reported) | Core |
| Creator-run Meta ads | Creators are both advertiser and payer; at least 5 creators | Reach only: Collin's ad reached 16,157 in the EU | Core |
| Google search ads | About 200 text ads, still running Nov 4 | SimilarWeb: 59% paid keywords; Paid Search is the #2 source | Core (capture) |
| Paid X creator quote posts | Several posts carry X's "Paid partnership" label | Views only: 4.37M views / 1,199 likes; @ftcarpe 61K | Support |
| In-product referral / affiliates | Code includes two-sided invites and affiliate payouts; creator links carry /i/ codes | No data | Support (amplifier) |
| Bus ads / trade shows | 130 buses in SF; 4 trade shows (incl. TechCrunch Disrupt booth A7) | No data (Summer: small early budget) | Support |
| Events | LA creator night with 100 people, free Pro on site | About 200-person event, 5M online reach (company-reported) | Support |
| PR wire releases | Business Wire / PR Newswire; The Block and GeekWire pieces look sponsored | No data | Support |
| App stores | No evidence of Apple Search Ads | Play 100K+ downloads; only 9 iOS ratings | Support |
| Meta brand ads | About 10 ads, started only on Nov 3 | EU reach of just 8 people | Pilot |
| ChatGPT ads | oaiq pixel installed; unknown whether ads are running | No data | Pilot |
| GEO | llms.txt about 31 KB, about 50 category terms; GPTBot and others allowed | No data | Pilot |
| Campus | Summer lists it as a pilot channel; $9.99 student price | Play reviews include student users | Pilot |
| SEO | Sitemap has only 10 URLs; blog stopped on Jul 11 | 41% organic keywords, mostly brand terms | Minimal |
| Product Hunt | Launched Nov 3 | #2 of the day, 367 votes, 200 comments; sign-ups unknown | Minimal |
| Owned social | IG / TikTok / YouTube / Discord / X all active | Few followers on every account (see section 6) | Minimal |
2. Which countries Google search ads target
~200Ads pointing to teamily.ai Verified
91Ads in the last 30 days
Nov 4Last shown (still running)
Search~200 ads; YouTube 24 / Play 0 / Shopping 20 (rechecked Nov 5)
Key points
- India has the most ads, the US is #2, and Germany, France, and Italy also have many.
- SG / MY / JP / TH / MX / CN / AE all have 0.
- Two advertisers: Teamily Inc (US, about 200) + TensorOpera HK Limited (Hong Kong, 21).
- All 8 sampled ads are text ads.
Source: Google Ads Transparency Center, teamily.ai. Ad counts aren't spend; one ad can target several countries, so countries overlap.
3. Where website traffic comes from Third-party estimate
617KVisits in the last 3 months, +31% vs prior period
59% / 41%Search keywords: paid / organic
81.3%Direct (paid tier, 6 months), #1; Paid Search #2
About 70% malePaid tier: 29.5% female
Top search terms
teamily ai · teamily · python-pptx · free ai video generator · teamly aiSocial referrals: YouTube first, then Instagram and WhatsApp.
SimilarWeb paid tier, Apr–Sep (US at 85.9% in the last 28 days). India has the most Google ads, but about 80% of site traffic is in the US: ad counts don't map to actual spend or traffic.
4. Site tracking: who they're paying
Pixels / tags Verified
Google Ads AW-18110929597 + GA4 + GTM Meta Pixel 2306850063429163 OpenAI Ads oaiq (bzrcdn.openai.com)// rareAdSense account tag// may be monetizing with adsPostHog product analytics// No AppsFlyer / Adjust / Branch (web only checked)
Attribution parameters (first + latest source stored in a cookie)
gclid wbraid gbraid// Metattclid// TikTok (possibly just reserved)msclkid// Bing (possibly just reserved)oppref obref utm_*
Bottom line: Google Ads is definitely being paid for Verified. Meta and ChatGPT pixels are installed, but whether ads run there is Inferred. TikTok / Bing only have parameters, with no evidence of ads.
5. Meta ads: mostly creators running ads from their own accounts
| Account | Followers | Dates | Region | Hook (comment keyword) |
|---|---|---|---|---|
| Collin Gourley @shot_by_cg | 3.7K | Oct 13–15 | 26 EU countries, reach 16,157 | "Comment TEAMILY"; 8 versions; an ad agency uses 3 agents to analyze client Meta ads |
| HackwithAntonio | — | Oct 15 | — | "Comment 'LOOP'"; 4 agents run a content pipeline; lands on teamily.ai/i/JMNSK4 |
| mehdi.ghazzai (Systemegagnant.fr) | IG 60.8K | Oct 29 | France | "Commente GAGNANT" (French) |
| Tomas Milinavicius | — | Sep 15–20 | — | — |
| Dimytechhub | — | Sep 5 | — | — |
| For comparison: brand page Teamily AI | 44 | Nov 3–4 | Worldwide 18–65+, EU reach 8 | No hook; the same 0:38 video, about 10 ads |
Sources: Meta Ad Library (Collin), creator ad, brand page. Each creator ad ran only 1–5 days (many for "23 hours"), as vertical UGC talking-head videos with #ad. Teamily pays, creators run the ads themselves, and branded Partnership Ads aren't used Inferred.
Creator funnel Verified
Sees creator video#ad, talking through one agent use case
→
Comments keywordTEAMILY / LOOP / GAGNANT
→
Gets link by DMteamily.ai/i/XXXX (with invite code)
→
Signs upInviter earns credits; attribution recorded at the same time
Summer says one UGC campaign had 150K+ people ask for the link in comments within a month (company-reported). The invite-code link does double duty for referral rewards and affiliate attribution.
6. Official accounts vs creator distribution
250M+Claimed monthly social views
≈15KCombined followers on five official accounts (1,403 + 2,341 + 279 + 467 + 10,300)
Reach comes from creator distribution, not official accounts. The official accounts are just landing spots. Inferred
The pinned shareholder letter on X has about 1M views; everyday posts get about 11K. No meaningful Reddit discussion; no official accounts found on Xiaohongshu, Douyin, or Weibo.
The pinned shareholder letter on X has about 1M views; everyday posts get about 11K. No meaningful Reddit discussion; no official accounts found on Xiaohongshu, Douyin, or Weibo.
7. App stores
A big gap between the two stores
100K+Google Play downloads; rated 3.6–3.7, about 450 reviews
9iOS ratings; rated 4.6, English only, not featured
- Long-time Play users say they "used it for 1–2 years, before the redesign" → the package predates the 2026 rebrand Verified
- The 100K+ includes old accounts and doesn't match "4.5M users" Inferred
- iOS reviewer ElleryWang999 may be a team member Inferred; Chinese-language reviews ask for WeChat/Alipay
- No evidence of Apple Search Ads
Countries with higher Productivity rankings (SimilarWeb)
| Android | Rank | iOS free | Rank |
|---|---|---|---|
| Thailand | #287 | Nigeria | #82 |
| Malaysia | #580 | Taiwan | #84 |
| Spain | #599 | Kenya | #148 |
| Mexico | #717 | Hong Kong | #230 |
| US | #1,137 |
iOS Taiwan at #84 is consistent with Summer's claim that Taiwan rose to #3 by users.
08 · Features
The pitch is humans and agents working together, but users come for research, slides, video and websites; marketing volume runs opposite to usage.
Features: launch → output → audience → marketing
The pitch is "humans and agents working together." Users actually use it for research, slides, video and websites. Marketing volume runs opposite to the usage ranking.
21features (incl. 1 non-feature story)
21/21with a launch date: high 10 · medium 9 · low 2
61% vs 14%of content on collaboration / social vs the top 4 output features by usage
237 : 18Jul–Aug changelog entries : official posts
Conclusions
- Marketing talks about collaboration and social; users use it for output. By this study's tagging, group chat, AI twin, multi-agent, memory and meetings make up 61% of feature-tagged content. The top 4 output features by usage (research, slides, video, websites) get only 14%. Video is #3 in usage but has 22 posts; slides are #2 with 43. Verified
- "Ship it, then post about it" held in only two windows. In Apr–May, the official X account posted 50 feature posts (almost one feature a day). In Jul–Aug, the changelog had 237 updates but only 18 official posts. Of 112 feature updates, 30 had an official post on the same feature within 7 days. Verified
- Creators picked up concept + scarcity fastest: Humans + agents in one group chat, Multi-agent teams / Agent Flows, AI twin / family and friends social, Living Memory / global memory got creator content in launch week (mostly beta reposts in February asking for invite codes). Real output features got creator coverage slowly: Local coding agent 45 days; Website builder / Web App Studio 53 days; Research / deep research 68 days. Verified
- The marketing agent and sales / ops are not launched features. From October they are use-case packaging aimed at small businesses and ad agencies, with no separate release notes, pushed through creator ads (3 agents analyzing a client's Meta ads, the LOOP content pipeline). Inferred
- Each audience gets a different hook (the feature with the most posts): creators → Multi-agent teams / Agent Flows, Research / deep research; office workers → Humans + agents in one group chat, Multi-agent teams / Agent Flows; students / side hustlers → Website builder / Web App Studio, Invite codes / two-sided referrals / referral credits. Small business owners show up in only 58 posts, though they are the company's stated main audience. Inferred
Launch timeline: ◆ launch · ▲ media · ■ release update · ● official post / creator content (size = views)
x = date, y = feature (sorted by launch date). Hover a dot for its title; use the slider below to zoom. Creator content clusters in February and November: February is the beta with invite-code requests, November is the widely shared shareholder letter. Undated Instagram / TikTok content is not shown.
Marketing volume vs actual usage
Bar = posts matched to the feature (official / creator / ads). Purple label = position in the shareholder letter's usage ranking; rank only, no volumes.
Feature × audience (posts)
Audiences come from content tagging (taxonomy_icp). "AI enthusiasts" are mostly beta-period reposts, not paying users.
Feature table Click a header to sort. Hover a feature for notes. All columns: data/feature_master.csv
Launch date = first public appearance described as usable (official X / media / changelog / iOS release notes / third-party posts). Confidence: high = official source; medium = media or indirect description; low = third party or use-case packaging. Lag = first marketing − launch (days); negative = marketing before a confirmable launch, or rule-tagging noise. Numbers are company-reported or third-party collected.
09 · Content
Micro / UGC accounts posted the most (62% of 1,165 posts), but owned accounts got 59% of all views.
Social channels and content: who posts, to whom, about what
Of 1,165 posts, the Micro / UGC channel has the most (62%). The Owned channel gets 59% of all views.
Scope: public content collected for this study (not a full census). Views and engagement are the values platforms display. Owned follower counts are from profile pages on Nov 5, 2026. Post counts Verified. ICP, use case and feature are tagging results Inferred. Sources: data/content_tagged.csv, data/creator_content.csv, data/creator_roster.csv, data/creator_stats_direction.csv, data/creator_stats_market.csv, data/creator_stats_weekly.csv, data/own_accounts.csv, data/trends_charts.json.
1. Channel overview: which platform, who posts
Right chart: creator reach = sum of followers of every KOL / KOC / micro account that posted on that platform (accounts deduplicated, audiences not). Log scale.
2. Creator pyramid: KOL / KOC / micro
Market × tier
Tier × content direction / ICP
3. Who says what to whom: channel × ICP / use case / feature
Cell number = posts. Hover to see total views and the top post. Click a cell to open that post.
4. Flow: platform → channel type → ICP
Line width = posts (not views).
5. By stage: what each stage talks about, and to whom
By stage = share (100% stacked; n on top = posts in that stage). By week = post count; undated posts are left out. Stages: 0 ≤ Dec 2025 · 1 Jan–Apr · 2 May–Jul 7 · 3 Jul 8–Oct · 4 ≥ Nov 1.
6. Which feature's content gets more views
Bar = mean views per post; dot = median. Gray bars have n<5; don't draw conclusions from them. One viral post can lift the mean, so check the median.
7. Content volume vs search interest
By creator tier (KOL / KOC / micro)
8. Top 20 by views
9. Creator roster
Click a header to sort. One row = one account × platform.
10. How to read this
What the data says
- Sample: 1,165 posts from 499 accounts on 12 platforms. 921 posts show a view count. Verified
- Most posts: Micro / UGC (727 posts, 62%). Most views: Owned (59%). Verified
- Views concentrate on X (89%). Verified
- Most targeted ICP: Office teams / employees (303 posts). Most common use case: Team collaboration / group chat workflow (424 posts). Verified
- Highest average views per post by feature (n≥5): Multi-agent teams / Agent Flows, mean 55,404, median 2,714 (n=160). Verified
- Creator tiers (accounts / posts): KOL 134/154; KOC 100/132; Micro / UGC 251/727. Verified
- 160 posts (14%) have ICP "Unclear". Read that heatmap column with care. Verified
Inferred Inferred
- X builds buzz and sells the concept (about 70% of KOL/KOC posts are company narrative); Instagram and Meta ads drive conversion (side-hustle money + comment-to-get-link); YouTube carries reviews and tutorials.
- Most "micro/UGC" content is organized: templated X replies and #teamilyaipartner micro-accounts posting per piece — not organic word of mouth. The highest views per post come from creators showing a specific job use case.
10 · PR
PR is credibility material for investors and AI search: global coverage is almost all paid wire releases, and search didn't move after any wave.
PR and media: what ran, where, and why
PR is credibility material for investors and AI search, not a user channel. Almost all global coverage is paid wire releases, and search did not move after any wave.
106 articlesChinese 46 · global 60 (reprints included, one row per article)
72% / 11%Global / Chinese clearly paid (wire + sponsored + contributor). Earned coverage only 18% / 43%; another 43% of Chinese articles look arranged
98 outletsMedia and channels (Chinese 46 · global 52). Zero original stories from TechCrunch, VentureBeat, e27 and similar
35 + 7Articles that link to teamily.ai (global + Chinese). Semrush Authority Score is only 20
All figures (users, funding) are company-reported and repeated as-is by outlets. Type rules: pages marked SPONSORED / paid press release / contributor = clearly paid Verified. “Suspected arranged post” = no ad label, but same-day release, shared press kit or reused sentences Inferred.
① Timeline: every wave across three regions, with “teamily ai” search
One dot = one original article; bigger dots have more reprints (the Korean syndication counts as 75 sites). Click a dot to open the article. Gray line = Google Trends “teamily ai”, worldwide weekly index (from Nov 2025, right axis). After Feb 13 search moved only slightly (weekly peak 17), it was 0 after May 26 and Jul 8, and it dipped in the week of Nov 3. Search reached 100 in October, when there were no new wire releases (that period had creators, UGC and search ads). Verified
② Each wave: channel, angle, numbers, purpose
| Wave | Date | Channel / wire service | Outlets | Angle | Claimed numbers | Purpose |
|---|---|---|---|---|---|---|
| ① ChainOpera crypto phase | Dec 2024 – Oct 2025 | The Block SPONSORED ×3; crypto newsflashes | The Block, SiliconANGLE, LABJ, PANews, Odaily, 深潮 (TechFlow), Cryptonomist | Funding / growth data | $17M seed; 1M+ DAU, 150K+ paying; 2M users | Token and funding story (the subject is ChainOpera, not Teamily) Inferred |
| ② Chinese + English launch | Feb 12–13, 2026 | ACCESS Newswire (EN); one shared Chinese press kit | 量子位 (QbitAI), 机器之心 (Synced), 36氪·智能涌现 (36Kr Emergence), 划重点 (KeyPoints); Yahoo Finance; Feb 15: 硅星人 (Silicon Star), 晚点 (LatePost) | Category first / founder story / WeChat comparison | $20M raised; “previous version had 3M sign-ups” | Fundraising: 36Kr quotes “a new round planned for March”; endorsement in Chinese VC/FA circles Inferred |
| ③ Contributor post + second push | Mar 6 / Mar 12, 2026 | Forbes contributor + podcast; 新智元 (AI Era) | Forbes (Charlie Fink), AI/XR Podcast; 新智元 (AI Era) and reprints | Review / riding the “WeChat AI” news | 10K+ beta users; “a year ahead of WeChat” | Re-exposure during the funding window; US credibility Inferred |
| ④ Going global: Singapore | May 26, 2026 | PR Newswire APAC | Straits Times (paid section), Manila Times, Tirto, Kosmo, Thailand Business News | Going global / national AI strategy | 4.5M+ sign-ups; US$20M seed | Government / channel endorsement; no local tech outlet followed up Inferred |
| ⑤ Official launch | Jul 8–27, 2026 | Business Wire multilingual + Simplified/Traditional Chinese; Korea Newswire | Yahoo, StreetInsider, MarTech Series, 中央社訊息平台 (CNA PR Wire), Yahoo奇摩 (Yahoo Taiwan), ~75 Korean local sites | Product launch (ICP shifts to startup teams) | Product stack 5M+; 20K seed users from public beta, hundreds paying | Launch credibility + one-off backlinks Inferred |
| ⑥ Fall piece | Oct 28, 2026 | GeekWire Contributor Content | GeekWire | Industry trend / big-tech comparison | Teamily 200K; TensorOpera 5M | Credibility material Inferred |
| ⑦ Growth announcement | Nov 3, 2026 | Business Wire + PR Newswire + PressAdvantage→XPR→Gannett; Product Hunt the same day | Yahoo, AFP, FinancialContent, AETOSWire, Akron Beacon Journal, Des Moines Register | Growth data / gross margin / funding comparison | 10K→320K sign-ups (32×); 20K DAU; positive gross margin | Pre-fundraise growth update, written for investors Inferred |
Wire services stepped up each wave: ACCESS → PR Newswire APAC → Business Wire multilingual + Korea Newswire → three wires on one day. Verified Global releases only tell the FedML → TensorOpera → Teamily story and never mention ChainOpera; the February Chinese articles don't either. Verified
③ How the user count changed (company-reported)
Log scale. Purple = numbers from the legacy user pool / product family (ChainOpera, TensorOpera); green = Teamily's own numbers. On Jul 8 the same BW release gave both 5M and 20K; on Nov 3 the figure became 320K. “US$20M seed” never changed. Verified The big numbers come from the old ChainOpera/TensorOpera pool. Inferred
The Block, Aug 2025, marked SPONSORED: ChainOpera 2M users
36氪 (36Kr), Feb 13, 2026: “previous version 3M sign-ups”
PR Newswire, May 26, 2026: 4.5M+ sign-ups
Business Wire, Jul 8, 2026: 5M+ and 20K seed users in one release
Forbes contributor, Mar 6, 2026: 10K+ beta users
GeekWire Contributor Content, Oct 28, 2026④ Outlet types and story angles
Left: by article (reprints included). Global: 37 wire releases, 3 sponsored, 3 contributor. Chinese: only Business Wire and 中央社訊息平台 (CNA PR Wire) are confirmed paid; none of the top AI outlets labeled an ad. Right: one article can have several angles.
⑤ Two signs this is not organic coverage
Duplicate-text check (Chinese) Inferred, strong evidence
12%
新智元 (AI Era) on Mar 12 shares 12% of its 14-character snippets with 量子位 (QbitAI) on Feb 13, including whole sentences and the closing line. It also stitches in paragraphs from 机器之心 (Synced, 4.4%) and 智能涌现 (36Kr Emergence, 2.9%). The same material was pushed again a month later. Verified
The four Feb 13 articles (量子位 / 机器之心 / 智能涌现 / 划重点) overlap only ~1% in body text, but use the same press kit:
- The same official post link x.com/teamily_ai/status/2022059522974757098
- The same English terms: Global Memory & Context Management / Social Brain Model / Agent Social Network
- The same team bio line: “worked at Apple, Amazon, Google, Tencent, ByteDance…”
- The same claim: “world's first AI-native social platform”
→ Same embargo date and the same press kit: a coordinated release.
Korean wire backlinks Verified
75 / 100
Of the top 100 referring domains in Semrush, 75 are Korean local sites from Korea Newswire syndication (same IP range 121.254.217.x, ~15 backlinks each). They first appeared 7/13–8/2, right after the Jul 8 BW release.
Month last seen: Aug 14, Sep 33, Oct 27, Still live in Nov 1. 74 were gone before November (Semrush marks 31 as lost). Wire backlinks are temporary and add authority only briefly.
Screenshot: Semrush referring domains
⑥ All articles
| Region | Type | Date | Outlet | Title | Angle | Article | Archive |
|---|
Archive = article text saved under report/raw/pr/. Blank = page is offline or blocks scraping; only search results remain as evidence.
⑦ Conclusions
- PR does not bring users. After Feb 13, “teamily ai” search moved only slightly (peak 17), was 0 after Jul 8 and dipped in the week of Nov 3. Search and web traffic stepped up twice, in August and late October, in line with search ads and creators / UGC. Verified
- Global PR is mostly bought. Of 60 global articles, 43 are wire, sponsored or contributor posts. No major tech outlet or AI newsletter wrote an original story. Top Chinese AI outlets shared one press kit on the same day, which looks arranged. Verified / Inferred
- The main goal is fundraising. The February Chinese articles sit right next to “a new round in March”. The November growth release covers positive gross margin and unit economics, and uses SimilarWeb to compare itself with peers that raised $100M–400M. The reader is an investor. Inferred
- The second goal is feeding AI search. Dozens of matching wire releases plus directories and AI-rewrite sites give ChatGPT / Perplexity many consistent lines to quote when asked “what is Teamily”. This matches the llms.txt setup. Inferred
- Shifting numbers are a risk. One company went 2M → 3M → 4.5M → 5M+ → 200K → 320K, which anyone can spot by comparing. Numbers should consistently describe either the product family or Teamily itself. Verified
Data: pr_coverage_cn.csv (46), pr_coverage_global.csv (60), pr_outlets_*.csv, semrush_backlinks.csv, trends_charts.json. Rebuild: python3 -I data/pr/build_section10.py --lang en
11 · Lessons
Copy the tactics: search ads as a demand test, one message per market, comment keywords for attribution. Not the numbers or the hype.
Lessons for AI product teams
A few specific tactics are worth copying: use search ads to buy demand signals, change the message by market, and attribute each creator with comment keywords. The numbers and the hype around them are not.
A sequence to try
1. Test keywords with search adsCompetitor names + broad "free ai xx tool" terms. Run small campaigns per country and see which keywords bring paying users.
→
2. English review channelsFind YouTube AI tool reviewers and put a unique invite code in the description.
→
3. Pick 1 low-CPM market1 local mid-tier paid creator + a few micro accounts, with a local message.
→
4. Re-rank weekly by paidSign-ups are just curiosity. Add budget where people pay.
Worth copying
| Tactic | Teamily's evidence | How to apply it |
|---|---|---|
| Search ads on competitor + free-tool keywords | Est. $280K spend in Aug, monthly visits ×11; keywords include perplexity ai, timely ai, free ai video generator Verified. Ads sell "AI PPT / Doc Generator — Free", not a concept | Write ad headlines about a concrete output ("Make an XX in one sentence"), not a platform concept. Test keywords on a small budget first, and keep them based on paid conversion, not visits |
| English YouTube reviews | About 38 videos, about 780K visible views, 21 with /i/ invite codes; the largest creator channel by views Verified | Find AI tool reviewers (50K–500K subscribers), give each one an invite code, and track sign-ups and payments by code |
| Comment-to-get-link | gptsavyy got 7,738 comments, kats.workspace 5,378; kiwi4work uses a new keyword per post to tell which one drove results Verified | One keyword + one invite code per video, with auto-DMs via a tool like ManyChat. Treat comment counts as a demand signal |
| One message per market | Taiwan: "free websites / don't want to go to work / an AI LINE group"; IG: side-hustle income; YouTube: "I built an AI team"; official channels: small business Verified | Write 2–3 hooks for the same product by market and platform, run them separately, and track conversion separately |
| Low-CPM market + one local creator | In Taiwan, Bagel alone posted 23 times for about 1.53M views; share went from 0.16% in Sep to 3.29% in the last 28 days Verified. Summer says total spend was <$20K (company-reported) | Pick one Chinese-speaking or Southeast Asian market, start with 1 local mid-tier creator, and check whether search and sign-ups move with it |
| Re-rank markets by paid conversion | Summer: "Sign-ups showed us where curiosity was" Verified (quote) | List sign-ups → paid by market in the weekly report and reallocate budget every week |
| Break out Direct traffic | The product runs on the main domain; Direct was 93% in Oct, mixing product usage, DM links, and likely bot traffic | Measure the marketing site and the product on separate domains or paths; tag every landing link with UTMs + an invite code |
Don't copy / discount heavily
- Shifting user counts: 2M → 5M+ → 320K falls apart as soon as anyone compares them. Report one definition only.
- Paid X quote posts + low-follower retweet farms: they buy views, not sign-ups (X ≈ 0 in SimilarWeb social referrals) Inferred.
- Paid wire releases as acquisition: neither the Jul 8 nor the Nov 3 release caused a search spike, and Korean wire backlinks dropped off within 2–3 months. Use wire releases only for credibility and fundraising materials.
- Broad keywords that buy low-intent sign-ups: South Asian free-tool keywords and Taiwan's "free website" hook raise sign-ups but lower the paid rate (about 1–2% paid) Inferred.
- Traffic with no clear source: Italy's 11% in Aug had zero dwell time, and steps switched on and off abruptly. Watch bounce rate and time on site, not just visits.
- Investing in official accounts first: median views on its official TikTok / YouTube are just 175 / 126. The reach is all in creators and ads.
Still unknown
| Question | Why it matters |
|---|---|
| Where the US traffic in the Oct 21 step came from | 1.16M US visits in the last 28 days; Taiwan and India explain only a few percent. It could be ads, product usage, or bots |
| Sign-ups and payments by channel | No channel has sign-up numbers, so causality can only be inferred from timing |
| Google spend in Oct | SimilarWeb only goes through Sep; Semrush only captured "ai agent" in India |
| Creator rates and total budget | The only figure is Taiwan's company-reported "<$20K" |
| Retention, paid rate, ARPU | These decide whether this acquisition model pays back; the only reported figures are "several thousand paid users, 20K DAU" |
| Real volume on TikTok and the DM funnel | TikTok scraping was blocked; the "150K people commented for the link" claim is about 7× higher than visible data |
12 · Evidence
Every key number and tactic is backed by a screenshot of the original post, release or ad library; figures are company-reported unless marked SimilarWeb.
Source screenshots
Every key number and tactic can be traced to the original posts, wire releases, or ad libraries. All numbers are company-reported; the only third-party data is SimilarWeb estimates.
Screenshots taken Nov 5, 2026. Yellow highlights were added when capturing. X posts are shown in the original language (no browser translation); the @liambraus post uses X's built-in Spanish-to-English translation. Colored outlines were added for this report: orange = company-reported, blue = third-party data, green = directly verifiable fact. Click an image to enlarge; "Original" opens the unedited capture.
Growth numbers
CEO post: "10K → 320K customers in one quarter, 100+ countries, 100M+ messages"; 909K views / 1.2K likes (the image is a counting animation; the captured frame shows 10,000) Verified
Shareholder letter: "10K → 320K sign-ups, 32×, 217.5% monthly compound growth"; claims to have passed Town/Viktor/wabi/cue on SimilarWeb; the post has 1M views Verified
Shareholder letter: "several thousand paid users, 20K DAU, 10K company teams, about 35% of 100M messages AI-generated"; offers enterprise private deployment Verified
Shareholder letter: 6 agent tasks per human-agent group per day + chart of top tasks (deep_search first) Verified
Opening of Summer's long post: 32× sign-ups, visits up nearly 15× Jul–Sep, 1M+ in Oct, trending twice in 9 days, 250M social views in a month, about 10K teams Verified
Playbook (from Summer's long post)
Tested paid/creators/UGC across 20 markets; ranked by sign-ups first, then re-ranked by paid conversion and "the map changed"; focused on 5–10 markets after a month; Taiwan went from outside the top 50 to #3 on <$20K, thanks to low CPMs and low AI fatigue Verified
9 channel types in parallel in month one (creators/UGC/paid/SEO/GEO/offline/events/campus/referral); small early offline budget; weekly reallocation Verified
Winning creator/UGC hooks → moved into paid → paid data feeds the next creator brief; one UGC campaign had 150K+ people ask for the link in comments within a month Verified
Attribution: sees video → sees it again → searches brand → signs up; last click goes to search / paid search Verified
User count definitions (ChainOpera)
The Block, Aug 20, 2025, labeled SPONSORED: "ChainOpera has 2M users and 50K developers, a WhatsApp-like agent social network" Verified
Source: theblock.co · Original
36Kr, Feb 13, 2026: "The previous version had 3M registered users; the new version builds seed users through invite codes"; pricing free / $19.9 / $199.9 Verified Inferred: the 3M are previous-generation (ChainOpera) accounts
Business Wire, Jul 8, 2026: the same release says the "product stack serves 5M+ users" and "about 20K seed users signed up in the past three months, with a few hundred paying customers" Verified Inferred: the 5M reuses the old ChainOpera user pool
Source: businesswire.com · Original
Ads and offline
Ads on 130 buses ahead of SF Tech Week (Golden Gate Transit bus wrap); the post has only 5,961 views Verified
2.0 launch post: "150K small-business owners and employees use it," "36M small businesses in the US"; 4.3M views vs 1.2K likes Verified Inferred: unusual view-to-like ratio → likely paid promotion
Homepage hero positioning: "The Human–Agent Platform…"; primary CTA is download for macOS / Web Verified
Source: teamily.ai · Original
Current pricing: Free $0 / Plus $19.99 / Premium $99.99 / Pro $199.99 (monthly; adds a $99.99 tier since the Feb coverage) Verified
Source: teamily.ai · Original
Ads and traffic
Google Ads Transparency Center: about 200 ads pointing to teamily.ai; advertisers Teamily Inc + TensorOpera HK Limited (previews didn't load) Verified
Source: adstransparency.google.com · Original
Website source (yellow highlights added when capturing): gtag AW-18110929597, OpenAI oaiq pixel, Meta Pixel fbq init 2306850063429163 Verified
Source: teamily.ai source · Original
Meta Ad Library search for Teamily: about 36 ads; creators running ads from their own accounts (mehdi.ghazzai, Collin Gourley, Tomas Milinavicius); Wax&Art ads land on *.teamily.run Verified
Source: facebook.com · Original
Brand page "Teamily AI" ads started Nov 3; advertiser and payer are both TENSOROPERA Verified
Source: facebook.com · Original
Creator Collin Gourley's ad: "Comment TEAMILY", #ad, ran Oct 13–15 across several EU countries, EU reach 16,157 Verified
Source: facebook.com · Original
Same ad: advertiser = payer = Collin Gourley (the creator runs it himself) Verified Inferred: Teamily pays the creator, who runs the ad
Source: facebook.com · Original
French creator mehdi.ghazzai's ad "Commente GAGNANT", ran Oct 29–30 for only 23 hours Verified
Source: facebook.com · Original
@nattcornelis quotes the CEO post with X's "Paid partnership" label; 50.8K views Verified Inferred: "trending twice in 9 days" likely relied on paid creator quotes
@liambraus Spanish post (auto-translated to English by X) quotes the CEO post with a "Paid partnership" label; 64.8K views Verified Inferred: "trending twice in 9 days" likely relied on paid creator quotes
SimilarWeb (third-party estimate): 617K visits in the last 3 months, +31.05% vs prior period Verified
Source: similarweb.com · Original
SimilarWeb: US 91.71%, Pakistan 2.7%, India 2.65%, Indonesia 0.85%, Egypt 0.81% Verified
Source: similarweb.com · Original
SimilarWeb: Direct traffic 84.29% (#1), paid search #2, organic search #3 Verified
Source: similarweb.com · Original
SimilarWeb: search traffic is 59.07% paid / 40.93% organic; top keywords are all brand terms (teamily ai / teamily) Verified
Source: similarweb.com · Original
The "Category Keyword Anchor Matrix" in llms.txt: tells AI which category term → anchor text → link to use (ai slack alternative, ai group chat…) Verified
Source: teamily.ai · Original
Top of llms.txt: entity statement "Teamily AI is a subsidiary of TensorOpera AI"; asks AI to use the full name when citing Verified
Source: teamily.ai · Original
US App Store: only 9 ratings (4.6), developer Teamily Inc Verified
Source: apps.apple.com · Original
Community engagement
Replies to Summer's long post: mostly short praise, each answered by Summer; accounts like liam., 泊舟 (en: Bozhou), and Naila overlap with paid promotion posts Verified Inferred: likely an engagement pod / paid creator group
