For SaaS and software companies

Six findings from the AuditLamp scan corpus

1

12.7% of 102 scanned business sites returned HTTP 403 to an AI crawler even though robots.txt explicitly allowed it. A firewall setting nobody had reviewed was overriding a policy nobody had written. For a SaaS company, this means a product site that appears open to crawlers may still be excluded from AI-generated tool recommendations because of a rule that nobody set intentionally.

Source: AuditLamp scan corpus, n=102, 2026-08-12.
2

Only 3 of 105 scanned business sites (2.9%) opened with a passage a machine could lift and quote as an answer. The qualifying passage had to be self-contained, specific, and attributable -- a sentence that stands alone without surrounding context. For a SaaS product, this means the homepage is almost certainly not structured to give an AI a quotable description of what the product does -- buyers asking AI for tool recommendations get a generic answer.

Source: AuditLamp scan corpus, n=105, 2026-08-11.
3

84 of those 99 failing sites had no concrete, citable claim near the top of the homepage. The first specific fact appeared after a hero, a tagline, and a list of feature names -- the worst possible structure for answer-engine extraction. For a SaaS homepage, a hero headline and a features list give AI nothing to quote when a buyer asks what the product actually does.

Source: AuditLamp scan corpus, n=105, 2026-08-11.
4

77.9% of the 653 reachable Tranco top-1,000 sites block no AI crawler at all. The sites that are open get read. The ones that are accidentally closed -- blocked by a default they never set -- do not know it. For a SaaS company, the products that appear in AI shortlists are almost universally the ones whose sites are fully open -- accidental blocking is a competitive disadvantage that does not show up in normal analytics.

Source: AuditLamp study, Tranco n=653, 2026-07-05.
5

Of 13 edge-level AI crawler blocks found across 102 scanned sites, only 2 were intentional. The other 11 were invisible -- a CDN default that nobody had reviewed and no robots.txt rule reflected. For a SaaS company, this means most AI-crawler blocking is not a privacy or competitive decision -- it is a default that no one reviewed and that no standard monitoring tool surfaces.

Source: AuditLamp scan corpus, n=102, 2026-08-12.
6

A study measuring what lifts content into generative-engine answers found that adding attributed quotations produced approximately 41% position-adjusted visibility lift. Keywords and formatting tricks did not produce the same result. For a SaaS product, pages that cite specific results, customer outcomes, or attributed statistics have a measurable advantage over pages built around positioning language and feature lists.

Source: GEO study (arXiv 2311.09735, KDD 2024), cited in AuditLamp scan note, 2026-08-11.

Can AI recommend your product?

AuditLamp is a free website scanner for SaaS and software companies. Enter your product URL and it checks 164 signals affecting whether AI systems and search engines can read and cite you: JavaScript rendering, AI-crawler access by name, answer extractability, and structured data. Free, whole site up to 150 pages, no email required.

Buyers build shortlists before you ever hear from them: a search, a "best tools for X" answer, a question to ChatGPT. The machines assembling those answers can only cite what they can read. If your docs and comparison pages are a JavaScript shell behind a bot wall, the shortlist gets built without you. Paste your link. Free scan. Worst problems first.

Whole site up to 150 pages · 164 graded checks · no email on the diagnosis · $10 full report if you want the PDF

The three failure modes

We look for the fails that lose evaluations.

1

Pages that go blank without JavaScript

Modern frameworks ship beautiful shells that show nothing until scripts run. Many crawlers never run them. The scan reads every page with zero JavaScript and flags the ones that disappear.

2

No passage a machine can lift into an answer

We check whether pages answer plainly near the top, whether structured data types the page correctly, and whether an answer engine gets a quotable sentence or a wall of positioning language.

3

AI crawlers blocked at two layers at once

We probe GPTBot, ClaudeBot, and PerplexityBot by name against your real robots file and your real edge behavior. Robots.txt can say yes while a firewall default says no. We wrote it up: your robots say yes, your CDN may still block AI.

Building AI products? There is a door for you too: for AI startups.

The honest boundary

Readiness, not a promised citation.

No tool can force a model or a search engine to shortlist you, and we do not pretend otherwise. The scan proves whether anything on your side blocks the machines from reading you, and ranks the fixes in plain language: 164 graded checks, worst problems first. If you want the file to keep, Full Report is $10 once. Ongoing re-checks are on pricing.

Check what AI and search engines actually find on your product site.

ai startups · ecommerce · designers · pricing