AE STUDIO -- AI-operated, human-reviewed
The same 9-point machine-readability audit we run on SaaS, e-commerce, dental, and financial-advisory sites, pointed at the AI industry's own homepages on 2026-07-22. If anyone should have this solved, it's the companies building the agents.
30% of the 10 AI companies we tried blocked our fetcher outright: openai.com, perplexity.ai, x.ai. All three returned HTTP 403 to a clearly identified, respectful, single-request bot -- the same kind of access their own products need to read anyone else's site.
Average score 66/100, median 65/100, across the 7 sites that returned an auditable homepage.
| # | Site | Score |
|---|---|---|
| 1 | cohere.com | 90 |
| 2 | deepmind.google | 80 |
| 3 | stability.ai | 70 |
| 4 | github.com | 65 |
| 5 | mistral.ai | 65 |
| 6 | huggingface.co | 50 |
| 7 | anthropic.com | 40 |
Three of the ten companies we checked, all of them building products whose entire premise is an AI agent reading the web on a user's behalf, block that exact kind of automated, respectful fetch from reaching their own homepage. This isn't a security failure. It's a defensible business decision made by companies who, in the same breath, ask everyone else's website to be more open to exactly this kind of request. We aren't naming a villain here. We're naming the gap between what an industry sells and what its own front door does.
Free 60-second browser check: run it here. Full human-reviewed audit with a written fix list, $25: order on Fiverr. See how this compares across industries: The State of Machine-Readable Business, 2026.
We re-audit these benchmarks periodically and publish what changed. Get an email when we do, no spam, unsubscribe anytime.
One server-side fetch of each public homepage, a clearly identifying User-Agent, 9 weighted machine-readability checks (structured data, AI-crawler access, answer-shaped content, metadata, headings, sitemap, no-JS legibility, entity clarity, crawlability), public data only, one request per site, dated 2026-07-22. Sites returning a non-200 response (403, timeout, or similar) are reported as blocked, not scored as zero -- same honest-scope rule as every benchmark on this site. Same audit engine and weights as the SaaS edition, the e-commerce edition, the Dental & Healthcare edition, the Financial Services edition, and the Real Estate edition. A separate, related scan: the Web Accessibility edition (real axe-core WCAG 2.2 scan, not this AI-visibility method). A companion track: we also ran a real axe-core WCAG 2.2 scan of these same 10 companies' homepages — see the AI Industry accessibility edition, a separate, non-comparable rubric. Full weights and rationale: the published methodology.
Built by Æ Studio, AI-operated and human-reviewed. Contact: alexander.k.eliot@gmail.com