46 B2B Tech Websites: AI Can Read Them, Not Cite Them

Most advice on AI search visibility opens with the same warning: ChatGPT and Perplexity can't see your site, because robots.txt or JavaScript is locking them out. I wanted to know how true that is, so I ran 46 B2B tech and SaaS homepages through my AI visibility check.

The result surprised me. On the technical side, these sites are ready for AI search: crawlers get in everywhere and the content is right there in the HTML. The problem is somewhere else. The pages don't answer questions, so AI can read them but finds very little worth citing.

Who I checked and how

I picked companies that sell expertise and implementation to other businesses, exactly the kind buyers now ask an AI assistant about ("who can implement our CRM", "a cloud provider for a mid-sized company"). All 46 come from the Polish market, split into five groups:

• CRM and ERP implementation partners (10 sites),
• IT outsourcing and consulting (8),
• cloud, infrastructure and security (10),
• data, BI and AI (8),
• B2B SaaS (10).

I checked each homepage once, on 17 October 2026, the way an AI crawler does: raw HTML without running JavaScript, plus robots.txt, llms.txt and the sitemap. The check has 20 tests in five groups and scores out of 100. The method and weights are on the check's page.

I'm not naming companies. This is a snapshot of the sector, not a ranking. The sample is hand-picked, so read it as a cross-section, not a representative statistic.

Overall: 78 out of 100

The median score is 78, the average 77, the range 44 to 98:

• 20 sites (43%) score 80 or more, which means they are technically ready to be cited,
• 24 sites (52%) land between 50 and 79, partly visible,
• 2 sites (4%) score below 50.

SaaS companies do best (median 85.5), cloud and security the worst (median 69). The other three groups sit close together, with medians of 77 to 78.5.

Access: not the problem

The most repeated fear didn't show up once:

• 46 of 46 sites let AI search crawlers in (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot),
• 46 of 46 serve their content in raw HTML; none is an empty app shell a crawler would see as a blank page,
• none carries an accidental noindex,
• all run on HTTPS and declare their language.

Only one company deliberately blocked the training crawlers (GPTBot, ClaudeBot, Google-Extended) while keeping the search crawlers open. That's a sensible setup: blocking training does not remove you from ChatGPT answers, which depend on OAI-SearchBot, as OpenAI explains in its crawler documentation. That site had the top score in the sample: 98.

llms.txt surprised me too. 24 of 46 sites (52%) publish one, more than have an FAQ. For context, llms.txt carries little weight in my check, because the major models haven't confirmed they use it.

One caveat: the check reads the rules in robots.txt, not firewalls. A firewall or CDN setting can block AI crawlers regardless of robots.txt, and that isn't visible from the outside.

Citability: this is the gap

The average score in the Citability group is 24%. The other four groups sit between 74% and 91%. That's the biggest gap in the whole study:

• 38 of 46 homepages (83%) have no questions and answers section,
• 43 of 46 (93%) have no FAQPage schema, the machine-readable version of those questions and answers,
• 22 of 46 (48%) don't have a single heading phrased as a question.

Why it matters: people ask AI in full sentences, like "how long does a Zoho rollout take" or "is staff augmentation worth it for a small team". A model looks for a passage that answers that directly. A tagline like "technology that drives business" tells it nothing. A question with a short, specific answer right below it is the easiest thing on a page to quote.

To be fair, FAQs often live on service pages, and I only checked homepages. But the homepage is the first page a crawler fetches when someone asks about the company itself. If all it finds is a slogan, AI has to guess, or take the answer from someone else.

Structured data: AI doesn't know who you are

The second gap is a missing statement of who is behind the site:

• 13 of 46 sites (28%) have no Organization schema at all, the JSON-LD description of the company,
• another 7 (15%) have one, but without sameAs links to LinkedIn, Clutch or the company registry,
• 12 sites (26%) declare no page type in their schema,
• 9 sites (20%) have no structured data whatsoever.

sameAs links look like a detail, but they're what lets a model connect your site to your profiles elsewhere. Without them, a company with a common name easily blurs into another one.

Content structure: small errors that add up

Finally, issues that don't kill visibility on their own but blur the message:

• 11 sites (24%) get the H1 wrong: six have none, five have several,
• 24 sites (52%) have headings that don't form an outline, with skipped levels or a single H2 for the whole page,
• 15 sites (33%) have a meta description that's too long, too short or missing.

The usual cause is the same: headings get picked for looks, not for content. There's an H3 because it looks nice, and the H1 is the logo.

My own site in the same check

I ran bergmann.studio too: 92 points. I was losing points on two things, as the homepage declared neither a page type nor an update date in its schema. I fixed both while working on this study. It took less than an hour, and the score went up to 98. The missing 2 points are for a generic WebPage type instead of something more specific, which I think is fine for a homepage.

What this means for your site

If your company looks like the ones in this sample, you probably don't need to fight robots.txt or rebuild your site on a different stack. You'll get more out of working on what AI finds once it's in:

1. Add 4–6 questions buyers ask before they buy to your homepage and service pages, with answers of 2–4 sentences.
2. Describe them with FAQPage schema, using exactly the text visible on the page.
3. Add Organization schema with sameAs links to your company profiles.
4. Clean up your headings: one H1 stating the page topic, H2 for sections, H3 for subsections.

How to check whether AI actually mentions your company in its answers is covered in my guide on checking your company's visibility in ChatGPT. The bigger picture, including content and brand authority, is in the AI search visibility guide.

FAQ

Does a high score mean ChatGPT will cite my company?

No. The score tells you whether AI can reach, read and understand the page. Citations also depend on whether your content answers real questions, and on how much other sources say about your company.

Why did you only check homepages?

Because it's the common denominator across all companies, and the first page a crawler fetches when someone asks about a company. Service pages are often in better shape, especially when it comes to FAQs.

Should I block GPTBot in robots.txt?

You can, if you don't want your content used to train models. Blocking GPTBot does not remove your site from ChatGPT answers; OAI-SearchBot handles those. Just don't block every crawler with a single rule.

How do I check my own site?

Enter your address in the AI visibility check. You'll get a score in seconds, no sign-up needed. If you'd rather have me review the site along with its subpages and rank the fixes by impact, get in touch about getting your site ready for AI search.