The lab notebook · near-daily
Lab notes
Short, dated, sourced. What changed in AI search, what the hourly Pulse caught, what the checker data shows — the working notes behind the evergreen playbooks.
Jul 20 hubspot.com scores 90/100 — strong visibility with minor tweaks needed hubspot.com scores 90/100, showing strong AI visibility with room for improvement in content structure. Jul 19 shopify.com at 80/100: AI-ready but needs schema work shopify.com scores 80/100, showing strong AI visibility but lacking detailed schema and concise answer structures. Jul 18 stripe.com at 100/100: strong visibility confirmed stripe.com scores a perfect 100/100 in our GEO Pulse probe, demonstrating strong visibility and AI readiness. Jul 17 anthropic.com at 65/100: visibility gaps to address anthropic.com scores 65/100 with partial visibility; key improvements needed in structured data and content mapping. Jul 16 openai.com at 20/100: largely invisible openai.com scored 20/100, largely invisible due to access issues and lack of structured guidance for AI crawlers. Jul 15 wikipedia.org at 65/100: partial visibility issues wikipedia.org scores 65/100, with AI visibility hindered by missing llms.txt and JSON-LD schema. Jul 14 notion.so at 73/100: partial visibility issues notion.so scores 73/100 with partial visibility due to malformed llms.txt and missing JSON-LD schema. Jul 13 github.com scores 80/100 — visibility strong but schema missing GitHub maintains strong visibility with AI engines but lacks JSON-LD schema, affecting entity recognition. Jul 12 Our own checker gave us 10/100 tonight. The bug class is probably on your site too. geoplaybooks.com scored 10/100 on its own AI-visibility checker: HTTP 522 on reachability, "no robots.txt found". The root cause — a worker cannot fetch its own zone — and why reachability failures ar Jul 12 We tried to review Scrunch AI. Its onboarding blocked us at the last step. Our hands-on Scrunch AI test ended at step 6 of 6: "This brand name violates our company brand policy." What we saw of the onboarding before the door closed — limits, auto-scraping, UX details — and t Jul 12 Profound pricing teardown (2026): $99/month buys 50 prompts and one engine We ran Profound's full onboarding to the paywall. Starter is $99/month for 50 prompts, ChatGPT only, one seat, no trial; Growth is $399/month for 100 prompts and 3 engines with a 7-day trial. There is Jul 12 Otterly.AI review (2026): eight minutes from signup to a real brand report Hands-on Otterly.AI review: 8 minutes from signup to a populated AI-visibility report, 6 engines tracked, sentiment and share-of-voice metrics, and the best PDF export of the four tools we tested — wh Jul 12 LLMrefs review (2026): one €79 plan, unlimited domains, and a free scan before the paywall Hands-on LLMrefs review: one All-in-One plan at €79/month with 500 tracked prompts, weekly runs, unlimited projects and seats, CSV + API — and a live free preview scan before you pay. What we saw, wha Jul 12 We ran our audit on 5 AI-visibility vendors. One passed. We pointed our 5-signal GEO checker at five companies that sell AI-visibility tracking. Scores: 100, 70, 55, 55, 40. Four fail their own medicine — and the failures are the same ones we see everywhere Jul 12 llms.txt adoption: the top 100 domains run 1.5x ahead of everyone else From our July 2026 crawl of the Tranco top 5,000: 11% of the top 100 serve a spec-valid llms.txt vs 7.4% for ranks 101–5,000. Adoption concentrates at the very top of the web. Jul 12 Checker Finds #1: robots.txt is never the problem. Schema is. Across the 13 domains our probes have logged so far, 13/13 pass robots.txt for AI bots — while 7 of 13 fail or half-fail JSON-LD schema and 6 of 13 stumble on llms.txt. The gap, in numbers. Jul 12 Vercel.com Achieves Strong AI Visibility with 93/100 Score Vercel.com is AI-ready with strong visibility, passing 4 of 5 core signals for AI-generated answers. Jul 12 We shipped an MCP server: the lab's tools, inside your AI agent GEO/AEO Playbooks now runs an MCP server: live 5-signal site audits, llms.txt linting, knowledge-base search and the Pulse benchmark — callable natively from Claude Code, Cursor or any MCP client. Jul 12 openai.com scores 20/100 on AI visibility. Here's the wall it hit. Our hourly GEO Pulse probe scored openai.com 20/100 on July 12, 2026: robots.txt allows every AI crawler, but the homepage answers HTTP 403. The robots-vs-WAF gap, explained.