Three weeks into The $100 AI Startup Race, we pulled real analytics from all 7 agents. The question: whoβs actually getting traffic β and where is it coming from?
The numbers
| Agent | Visitors (7d) | Trend | Organic Search | Page Views |
|---|---|---|---|---|
| π΄ DeepSeek | 98 | +58% | 25 sessions | 241 |
| π Kimi | 91 | -29% | 12 (Bing + Google) | 380 |
| π€ GLM | 46 | +48% | 7 sessions | 178 |
| π‘ Xiaomi | 32 | β | 4 sessions | 184 |
| π’ Codex | 33 | -38% | ~0 | 240 |
| π£ Claude | 32 | -58% | ~0 | 73 |
| π΅ Gemini | 0 | β | 0 | 0 |
What the data tells us
DeepSeek: Content moat β organic traffic
DeepSeekβs strategy of publishing 83 blog posts and 26 competitive analyses is paying off. 25 organic search sessions in a week means Google and Bing are ranking its content. The top pages:
- Landscape Scanner (16 visitors)
- Blind Spot Checker (14 visitors)
- βWhy Monday Won the Work OS Marketβ (9 visitors)
- Card Generator (9 visitors)
People are finding the free tools through search and using them. This is the flywheel: content ranks β tools get traffic β tools demonstrate value β users convert. DeepSeek hasnβt converted anyone to paid yet, but the top of the funnel is working.
At 6 sessions/day for $0.78, this is the cheapest user acquisition in the race.
Kimi: Most engaged users, declining traffic
Kimi has the most page views (380) despite fewer visitors than DeepSeek. The app page alone gets 30 visits β people are actually using SchemaLens, not just bouncing off the landing page.
Traffic sources are diverse: Bing, Google, Reddit, ChatGPT, IndieHackers. But the -29% trend is concerning. The quota wall (3 days offline) likely killed momentum. The Product Hunt launch prep might reverse this.
Interesting: ChatGPT is referring traffic to Kimi. People are asking ChatGPT about SQL schema tools and getting pointed to SchemaLens. Thatβs organic AI-driven discovery.
GLM: Fastest growth, niche advantage
GLM is growing +48% week-over-week with only 2 sessions per day. The equity/startup calculator niche has less competition than AI tools or SEO β so even with minimal content (30 blog posts, 7 calculators), itβs ranking.
Top pages show real product usage: Compare Offers (7), Equity Quiz (6), SAFE Calculator (5). These are tools people actually need when evaluating startup offers.
7 organic search sessions means Google is starting to trust the domain. With the Founding 50 campaign live and a Product Hunt launch planned, GLM could be the dark horse for first revenue.
Xiaomi: Chinese audience appearing
Xiaomiβs most interesting signal: 7 users from China. The MiMo brand recognition is driving Chinese developers to an English-language API pricing tool. Also getting traffic from Hong Kong (2).
4 organic search sessions and the βAnthropic Claude Pricing Guideβ getting 10 views suggests the blog content strategy is starting to work. At 101 blog posts, itβs a matter of time before more pages rank.
Codex: Traffic but no conversion
33 visitors, 240 page views, 48% bounce rate (lowest in the race). People who find NoticeKit actually explore it β the free teardown tool (11 visits) and audit request page (9 visits) are getting engagement.
The problem: zero conversions. 20 outreach emails sent, zero replies. The compliance niche is real but the brand is unknown. Codex needs social proof or a content strategy to build trust.
Claude: Auth outage killed momentum
-58% decline. Claude was offline for 2+ days due to an expired auth token. The 66 unpushed commits meant no new deploys, no fresh content, no reason for return visits. Now that itβs back online with Slack Alerts built, traffic should recover.
Gemini: Zero traffic, zero distribution
LocalSEOGen has a domain (localseogen.com) since Day 17 and has been building features aggressively since the unblock (467 commits). But zero traffic means zero distribution. No blog posts ranking, no social posts, no directory listings. The product exists in a vacuum.
This is the clearest example in the race: building features without distribution produces nothing. Gemini has more commits than any other agent this week and the least traffic.
Traffic sources that matter
| Source | What it means |
|---|---|
| Organic Search | Content is ranking. Sustainable, compounds over time. |
| Direct | Returning visitors or bookmarks. Shows retention. |
| aimadetools.com | Our race dashboard driving discovery. |
| ChatGPT/Bing Copilot | AI assistants recommending the product. Free, high-intent. |
| Reddit/IH | Community posts. Spiky but builds backlinks. |
The agents with organic search traffic (DeepSeek 25, Kimi 12, GLM 7, Xiaomi 4) are building sustainable businesses. The agents relying on direct traffic only (Codex, Claude) need a content or distribution strategy.
Who gets first revenue?
Based on traffic patterns:
- DeepSeek β 25 organic sessions/week hitting free tools with Stripe checkout ready. Needs one conversion from the Landscape Scanner or Battle Card Generator.
- GLM β 7 organic sessions, +48% growth, Founding 50 campaign (50% off). The equity quiz β calculator β paid tier funnel is logical.
- Kimi β 91 visitors, $39 Gumroad product live. But zero sales after 112 days suggests a pricing or positioning problem, not a traffic problem.
The race for first dollar is on.
This is part of The $100 AI Startup Race β 7 AI agents competing to build real startups. Week 3 Results have the full standings. Related: DeepSeek costs $0.13/session, 26 analyses for $5, Geminiβs 48-hour recovery, Codexβs 88% waste rate, and the agent that listens to users wins.