The state of AI readiness in SaaS: 100 leading homepages, scanned
We ran the Citation Readiness Score™ — the deterministic, 100-point Zitatix audit of how ready a page is to be found, quoted, understood, and trusted by AI answer engines like ChatGPT, Perplexity, Gemini, and Claude — against the homepages of 100 leading SaaS companies.
Study date: · Methodology version 1.2 · Fully reproducible: re-scan any listed homepage at zitatix.com to verify the numbers.
Headline findings
- 93% of leading SaaS homepages have no question-and-answer content — the single format AI answer engines most prefer to quote.
- 87% fail the content-to-code ratio check: their homepages are so JavaScript- and markup-heavy that the text an AI crawler can extract is a sliver of the page.
- 86% publish no machine-readable dates, leaving AI engines unable to tell whether the content is current.
- 29% ship no JSON-LD structured data at all — not even an Organization schema.
- Six household names scored below 50/100, including companies worth tens of billions.
- Blocking AI crawlers is rare: only 3% block GPTBot in robots.txt. Access is not the problem — extractability is.
The average score was 81.4/100 (median 86), but the spread is wide: 72 sites scored 80 or above, 21 landed between 50 and 79, and 6 fell below 50. The top performer scored 96; the bottom scored 18.
Why this matters
A growing share of product research now happens inside AI answers instead of search result lists. When a buyer asks ChatGPT or Perplexity "what's the best tool for X," the engines quote pages they can access, extract, and trust. This study shows that even the companies with the largest marketing budgets in software systematically fail the extraction and trust layers — so the citations go to third-party review sites and blogs instead of the vendors' own pages.
Method
- Sample: 100 well-known B2B and prosumer SaaS companies (CRM, dev tools, fintech, HR, martech, security, productivity). Homepages only, scanned July 16, 2026.
- Instrument: the Citation Readiness Score™ — roughly 25 deterministic checks across four pillars (Findable, Quotable, Understandable, Trustworthy), 25 points each. No AI model is involved in scoring; anyone can reproduce any score by re-scanning the same page. Full details on the methodology page.
- Coverage: 99 of 100 sites scored; one homepage exceeded the analyzer's page-size limit — itself a finding.
- Scope: on-page readiness only. Off-site reputation (the AI Visibility Index™) was deliberately excluded to keep the study 100% reproducible.
Results by pillar
| Pillar | Average (of 25) | What drags it down |
|---|---|---|
| Findable | 24.0 | Almost nothing — HTTPS, indexability, and sitemaps are near-universal |
| Quotable | 19.9 | 93% lack Q&A content; clever-but-vague headlines |
| Understandable | 19.2 | 87% fail content-to-code ratio; 56% have image alt-text gaps |
| Trustworthy | 18.3 | 86% lack machine-readable dates; 29% have no structured data |
The pattern is unambiguous: SaaS has solved findability and failed quotability. The classic SEO layer (HTTPS, sitemaps, indexability, canonicals) is essentially perfect across the sample. The AEO layer — content an engine can lift and trust signals it can verify — is where the points die.
The failure leaderboard
Share of the 99 scored homepages failing or only partially passing each check:
| Check | Fail or warn | Hard fail |
|---|---|---|
| Question-and-answer content | 93% | 5% |
| Content-to-code ratio | 93% | 87% |
| Machine-readable publish/update dates | 86% | 22% |
| Image alt text | 56% | 26% |
| Organization / author identity | 38% | 0% |
| Structured data (JSON-LD) | 29% | 29% |
| Meta description | 26% | 6% |
| Single clear H1 | 24% | 9% |
AI crawler access: mostly open
Contrary to the "everyone is blocking AI" narrative, robots.txt blocking is rare among SaaS companies: GPTBot is blocked by 3 of 99 sites, ClaudeBot and PerplexityBot by 1 each. SaaS wants to be in the answers — it just hasn't made its pages usable once the crawler arrives.
The spread
Top scorers (96–92) share the same traits: server-rendered content, real text on the page, structured data, and clean heading hierarchies. Bottom scorers (18–33) — several of them enterprise giants — share the opposite: homepages that are effectively application shells, with heavy client-side rendering, little extractable HTML text, no dates, and thin structured data. Enterprise scale clearly does not purchase AI readiness.
The three cheapest fixes
- Add machine-readable dates (86% affected) — an
article:modified_timemeta tag ordateModifiedin JSON-LD. Minutes of work. - Ship JSON-LD structured data (29% have none) — an Organization schema is table stakes; FAQPage where Q&A exists.
- Add a real FAQ section (93% affected) — answer, in two sentences each, the five questions buyers actually ask an AI about your category. This is the content engines lift verbatim.
Reproducibility
Every number in this study can be verified by re-scanning any listed homepage at zitatix.com — the Citation Readiness Score™ is deterministic and versioned (this study: methodology 1.2). Scores may shift as sites change; that is the point.
Press inquiries: info@zitatix.com