CrawlCheck

Findings · 2026-09-30 · By

Eight contradictions in our own copy, rechecked

An outside audit listed eight places where crawlcheck.io contradicted itself. Rechecked against the live site: which were real, which were two truths read as one, and the one still wrong.

On 25 September 2026 an outside audit listed eight self-contradictions in crawlcheck.io's copy. Rechecked against the live site, three were two true statements read as one false one and are now stated together, two were real and were fixed, two were consistent as written, and one was still false: a post claiming no scanned domain is ever named in anything CrawlCheck publishes, although the directory and registry name sites. That sentence is now scoped correctly.

Eight claims from the audit, and where each stands
1. The count of crawler identities differed between pages
What the copy says now
One figure everywhere, from one source
State
fixed
2. Measured sections and scored sections gave two different numbers
What the copy says now
Both numbers, said together
State
two truths, now in one place
3. Agency 'no daily cap' versus a monthly API quota
What the copy says now
The cap-free scanning is on the site; the quota belongs to the API product
State
two products
4. 'No scanned domain is ever named' versus a directory that names domains
What the copy says now
Scoped to the article and the dataset's aggregate figures
State
was false; corrected
5. Report 'regenerates' versus a retention period
What the copy says now
One sentence states both
State
two truths, now in one place
6. 'Every request identifies as CrawlCheck' versus fetches under named crawler user-agents
What the copy says now
The policy page lists every identity sent
State
fixed 23 September
7. Agency versus a larger plan called Network
What the copy says now
More than 40 domains: Network, quoted
State
consistent
8. App 'in build' versus an installable web app
What the copy says now
In build for iPhone and Android; no web app manifest is served
State
consistent as written

On 25 September an outside audit listed eight places where the site's own copy contradicted itself. Each was rechecked against the live site, not from memory of what had been fixed.

#The audit's claimWhat the copy says nowState
1The count of crawler identities differed between pagesOne figure everywhere, from one sourcefixed
2Measured sections and scored sections gave two different numbersBoth numbers, said togethertwo truths, now in one place
3Agency 'no daily cap' versus a monthly API quotaThe cap-free scanning is on the site; the quota belongs to the API producttwo products
4'No scanned domain is ever named' versus a directory that names domainsScoped to the article and the dataset's aggregate figureswas false; corrected
5Report 'regenerates' versus a retention periodOne sentence states bothtwo truths, now in one place
6'Every request identifies as CrawlCheck' versus fetches under named crawler user-agentsThe policy page lists every identity sentfixed 23 September
7Agency versus a larger plan called NetworkMore than 40 domains: Network, quotedconsistent
8App 'in build' versus an installable web appIn build for iPhone and Android; no web app manifest is servedconsistent as written

The one that was still wrong #

The post on what a crawler actually receives said no scanned domain is ever named in anything CrawlCheck publishes. That stopped being true on 10 September, when the directory and the registry began listing sites by name with their grades. The sentence now says the article and the dataset's aggregate figures name no site, and that the directory and registry do, by a stated method with a dispute route.

Why two true statements read as a contradiction #

Items two, three and five were each correct on their own page. A reader who meets them on different pages cannot tell they are compatible, so the audit's instinct was right. The fix was to put both halves in one sentence.

The two real ones #

The policy page said CrawlCheck identifies itself on every request. It does not: comparing what crawlers receive means fetching under their names. The policy page now lists the identities used. The identity count had drifted between pages as the list grew; it is now one figure wherever it is printed.

19

The lesson for your own site #

A number printed on more than one page will drift unless every page reads it from one source. A claim that begins never or every has to be checked against every feature, not only the one it was written beside. The rest of that week's changes are in the 29 September release notes.

Share this finding:LinkedInXEmailFollow Emmanuel on LinkedIn

Questions this post answers

How many of the eight contradictions were real?

Two were real and fixed, one was still false when rechecked and was corrected, three were two true statements read as one false one and are now stated together, and two were consistent as written.

Which claim was still wrong?

A post said no scanned domain is ever named in anything CrawlCheck publishes. The directory and registry name sites, so the sentence was scoped to the article and the dataset's aggregate figures.

Does CrawlCheck identify itself on every request?

No. To compare what different crawlers receive, it fetches pages under the user-agent strings of named crawlers. The policy page lists them.

How do you stop a number contradicting itself across pages?

Print it from one source everywhere it appears. A hand-typed copy of a changing number will drift as soon as the number changes.

Related findings

The main product

Found this on your own site? We fix it for $100.

Scan free to see where you stand. The fix is one site, every finding implemented and re-measured, with a sealed before and after.

Free browser extension

Check this on any site you open, without leaving the page.

CrawlCheck for Chrome grades every site you visit A–F and shows, from your own browser, what search engines and AI assistants are actually served. Free, no account.

Write for the library

Measured something yourself? Submit an article.

Send your finding, the data and the method. A person reads every submission and replies by email; accepted articles are published under your name with a link to your study. Free to submit. Articles are accepted on merit; no one can pay to have one published.

Submit an article

How it works · a reply within about a week

Spotted a mistake or a bug on this page? Report an error. Confirmed errors are fixed and logged on the corrections page.

Comments

Comments are read before they appear. Nothing is published automatically, and no account is needed. An AI agent with a published signing key can post without the queue: it signs the request and the signature is checked against its own domain (how). Those appear at once, marked verified.

Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.

All findings · The dataset · How the dataset works