AI search scores of YC Fall 2026, six weeks before Demo Day
153 YC Fall 2026 homepages, checked for how well ChatGPT and AI answers can read them. Median 72, seven perfect scores, and the leaders in each space.
Demo Day for YC's Fall 2026 batch is on 2 December. Between now and then, investors, customers and journalists will look these companies up, and more of them now do it by asking ChatGPT or Perplexity. So, six weeks out, we asked the same question we asked of the Summer batch: when an AI answer goes looking for a company in this batch, how much of its homepage can it actually read?
On 21 October 2026, YC's directory listed 158 Fall 2026 companies. We took the 155 that list a website and ran each homepage through our AI search check. 153 answered; two were down that day.
The scores
Median 72 out of 100. The Summer batch sat at 66, but we have tightened one check since then (more on that below), so treat the comparison as rough.
Seven perfect scores, and 22 companies at 90 or above.
17 under 50, and 13 homepages send 50 words of HTML or fewer before JavaScript runs, which is close to blank for many AI crawlers.
Score
Companies
90 to 100
22
80 to 89
32
70 to 79
27
60 to 69
29
50 to 59
26
Under 50
17
A batch homepage is usually a few weeks old, built to win the first customers, and it does that job. The point is that the same page is also how a model decides what the company is, and that decision is being made now, before Demo Day.
The leaders in each space
The top three in every space with ten or more companies, grouped from YC's own categories. Scores as of 21 October 2026; each links to a live re-check, so you can see where it stands today.
The seven perfect scores: DeepMark, Clean, Papaya, Mirrors, Vyra, Veeza AI and Databuddy. None of them needed a big budget. Each sends plain text that says what the product is, labels the company and the product in structured data, answers buyers' questions on the page, and has an llms.txt.
The widest gap between spaces is the same as in Summer: hardware. Industrials, the robotics, defence, energy and drone companies, sit at a median of 58.5, against 77 to 80 for the software spaces. Their homepages say much less in text: a median of 279 words in the HTML, against 614 across the batch.
What most of the batch is missing
No FAQ on the homepage: 64%. The questions buyers ask (price, who it is for, how it compares) go unanswered on the page.
Nothing that describes the product to machines: 70% have no Product, Service or SoftwareApplication data.
No llms.txt: 57%. It takes an hour to write.
Nothing that says who the company is: 46% have no Organization data, so a model has to guess which "Clean" or "Boat" is meant.
No sitemap: 24%. No canonical address: 22%.
Five fixes before Demo Day
Put the one-line pitch in the HTML, not only in a video or an animation. "X is Y for Z", in text a crawler can read.
Label the company and the product in structured data, with links to your LinkedIn, X and YC company page, so engines can tell you apart from a namesake. It's good practice that labels, not a guarantee of anything.
Answer five questions buyers ask, on the page, two sentences each.
Add an llms.txt and a sitemap. Two files, an hour together.
Re-check after every redesign. Batch sites change weekly, and so do these scores.
What this does and doesn't say
This measures how readable a homepage is to AI search, not how good the company is or whether ChatGPT recommends it. It is one page per company, checked on one day, and only the companies in the batch that list a website.
How we measured
What we checked
Every company in YC’s Fall 2026 batch that lists a website on the YC directory: 158 companies listed on 21 October 2026, 155 with a website, the homepage of each; 153 scored (two returned errors). The full list is below.
When
21 October 2026, one run per site.
With what
Our free AEO and GEO check, with its sixteen weighted checks as they stood that day (crawler access, text in the server HTML, structured data, answers to buyers' questions, titles and descriptions, llms.txt, sitemap, freshness and more) (the scoring code↗ (opens in a new tab)). It reads the HTML the server sends, without running JavaScript, plus robots.txt, llms.txt and the sitemap. Since 18 October the FAQ check counts questions a reader can see on the page, and since 20 October it ignores decorative icons after a question. Our Summer 2026 study used the earlier FAQ rule, so the two medians are not like for like.
Spaces
Grouped from YC’s own industry and subindustry labels: engineering, product and design tools and infrastructure on their own, the rest of B2B together, and Industrials, Consumer, Healthcare and Fintech as YC lists them. Spaces with fewer than ten companies are left out of the table.
What the score means
A score out of 100: the weighted share of the sixteen checks a page passes, with a partial pass earning a third of its weight. A page that blocks AI crawlers or asks not to be indexed is capped at 40. It measures how readable a page is to AI search, not whether any AI answer names or recommends the company.
Limits
One page per site, one run, one day; sites change, so a fresh check may differ. Every score named in this post links to a live re-check.
Per-site results
We name only top scorers in public. Each company can have its own full report from us; anyone can run any site on the check and see its score today.