We audited 110 SaaS homepages for AI search readiness
An original audit of what AI search engines can read on 110 SaaS homepages, from public giants to bootstrapped tools. The technical side is mostly handled. The writing is not.
- On the homepage 22
- Only in llms.txt 42
- Neither 40
- Refused our request 6
- publish an llms.txt file for AI tools
- 69%
- open their homepage by saying what they do
- 21%
- of 107 block any major AI crawler
- 2
- median homepage HTML, before a single image
- 536KB
The short version
We checked 110 SaaS homepages in October 2026 for the things AI search engines read. The technical side is mostly done: 69% publish an llms.txt file, 74% describe themselves with Organization schema, and only 2 of 107 block any of the main AI crawlers from their homepage.
The writing side is not. Only 21% of homepages open with a plain sentence that names the company and says what it does, and just 9% use the "Company is a..." form an answer engine can lift as it stands. Most of these companies explain themselves clearly in a file written for bots, then open their homepage with a slogan.
One dot per company
All 110 homepages, grouped by stage. Switch the check to see who passes.
69% publishes llms.txt: 75 of 109
21% opening says what it does: 22 of 104
74% has organization schema: 77 of 104
13% links to wikidata: 13 of 104
2% blocks an ai crawler: 2 of 107
Public 32
Late-stage private 36
Growth 24
Early or bootstrapped 18
The checklist is done. The copy isn't.
Share of the 104 homepages that loaded, for each check. The numbers fall off a cliff at exactly the point where the work stops being a checkbox and starts being writing.
The answer is there. It's just in the wrong file.
69% of these companies publish an llms.txt, and 75% of those files open with a clean definition. Their homepages mostly don't. Here is the pattern, with a made-up company.
Work, reimagined for the AI era.
Join 30,000 leaders at Acme Summit this June. Save your seat.
Get started# Acme
> Acme is a scheduling platform for
> physiotherapy clinics.
## Product
- [Pricing](https://acme.example/pricing)
- [Docs](https://acme.example/docs)Work, reimagined for the AI era.
Acme is a scheduling platform for physiotherapy clinics. It books, reminds and bills in one place.
Get startedWhere the definition lives
The 73 companies where we could read both the homepage and an llms.txt.
AI crawlers are welcome almost everywhere
What the 107 readable robots.txt files say about the four big AI crawlers. Most never mention them, so they inherit the general rules and are let in.
- GPTBot OpenAI 21% named Blocked by Figma and Canva
- ClaudeBot Anthropic 21% named Blocked by Figma and Canva
- PerplexityBot Perplexity 21% named Blocked by none
- Google-Extended Google 18% named Blocked by Figma
6of 110
The gap robots.txt doesn't show
6 homepages (5 behind Cloudflare, 1 behind Akamai) answered our clearly labelled research bot with a 403. Where we could read robots.txt, it would have let a bot like ours in, so the refusal came from bot protection at the edge. If your firewall is stricter than your robots.txt, the crawler you invited can still be turned away at the door. Check both.
Growth-stage companies write the clearest openings
The parts that need intent rather than budget are led by growth-stage teams. Public companies have the most complete schema and the least direct copy. Treat the splits as directional; the sample was picked by hand.
| Stage | Homepages | Direct answer | llms.txt | Organization | Wikidata | Median HTML |
|---|---|---|---|---|---|---|
| Public | 29 | 14% | 66% | 83% | 17% | 610 KB |
| Late-stage private | 33 | 18% | 69% | 76% | 18% | 757 KB |
| Growth | 24 | 38% | 83% | 75% | 8% | 392 KB |
| Early or bootstrapped | 18 | 17% | 56% | 56% | 0% | 267 KB |
Homepages are heavy
The median homepage sent 536 KB of HTML before a single image, script file or stylesheet. 15 sent more than 1 MB, and the largest sent 2.5 MB. A crawler reading raw HTML has to dig through all of it to find the two sentences that matter.
What this means for your homepage
If you run a SaaS site, the technical checklist is probably done. The cheap wins left are almost all in the copy.
- 1
Open with the definition
First paragraph: your company name, "is a", the category, and who it's for. Keep the slogan in the H1 if you like it, and say the plain version right after.
- 2
Reuse what you wrote for bots
If your llms.txt already has a good one-line description, it belongs on the homepage too.
- 3
Connect your entity
Add a Wikidata link to your Organization sameAs once you have an entry. Only 13% do.
- 4
Check your firewall, not just robots.txt
Make sure the AI crawlers you allow can actually load the page.
- 5
Watch the weight
If your homepage HTML is over 1 MB, find out what's in it.
How we did it
We picked 110 SaaS companies by hand to spread across stage and category: public companies, late-stage private, growth-stage, and early or bootstrapped, covering developer tools, data, security, productivity, sales and marketing, finance and HR, support, design, vertical software and AI-native products. It is a deliberate spread, not a random sample.
For each company a script made three requests, one at a time: the homepage, /robots.txt and
/llms.txt. No logins, no retries, no JavaScript, and every request named us in its user agent.
Skipping JavaScript is the point: most AI crawlers read the HTML a server sends and nothing more.
One snapshot, fetched on October 17, 2026 from India, so a few sites served regional versions of their homepage. "Opening says what the company does" is a heuristic: the H1 or first real paragraph must define the company ("Company is a...") or open a sentence with the company name as the subject of a verb, ignoring award and acquisition news. We spot-checked it by hand, but it will miss phrasing it doesn't expect.
