public pages only
no login, no install
Two of your own pages disagree about your product.
OfferTruth AI reads your public pricing, feature, docs and help pages, then reports only the contradictions it can prove.
Every finding carries two source URLs, both sentences quoted verbatim, the scan date, and the corrected text you can publish.
Your pricing page and your help center, printed and read next to each other. Almost nobody does this.
01 the finding
A finding is one object you can check in ten seconds.
Two claims, two live URLs, both sentences quoted exactly as they are written, and the sentence we suggest instead. You click your own link and see it yourself.
Trial length conflict
severity medium|confidence 96/100
"Start your 14-day free trial. No credit card required."
https://example.com/pricing · scanned 2026-08-24
"Your trial lasts 30 days from the day you create a workspace."
https://example.com/docs/getting-started · scanned 2026-08-24
Proposed correction
"Start your 30-day free trial. No credit card required."
Uncertainty note
The docs page carries no publication date. If a 30-day trial replaced a 14-day trial, the pricing page is the page to change.
Verified twice against live pages · 2026-08-24 14:02 UTC
Anonymized sample. In a real report every URL is clickable and every excerpt is verbatim, 40 words maximum.
02 why it is there
Nobody reads their own site end to end.
A 2–50 person SaaS company publishes its commercial promises in a dozen places: homepage, pricing, feature pages, integrations, docs, help center, FAQ, changelog, cancellation and refund notes.
They were written at different times by different people, and almost never reconciled. Four plan changes later the pages have drifted apart, and nobody inside the company sees it.
Pages that state the same claim type
/pricing
price · billing period · trial duration · seat limit
/docs/getting-started
trial duration · seat limit
/help/billing
refunds · cancellation
/features/integrations
integrations · plan eligibility
Three of these four pages state a trial duration. One of them is eleven months older than the others.
A prospect who notices asks support, quietly discounts the claim, or leaves. An assistant researching you reads both sentences and repeats whichever one it found.
03 what we compare
Eighteen kinds of claim, normalized and compared with each other.
Each public page is reduced to plain facts. Then every fact is checked against every other fact on your site.
- Price
- Currency
- Billing period
- Trial duration
- Trial conditions
- Free plan
- Seat limits
- Usage limits
- Storage limits
- Included features
- Excluded features
- Plan eligibility
- Support availability
- Cancellation
- Refunds
- Integrations
- Availability
- Version status
A disagreement is not yet a finding. Every candidate is tested against the innocent explanations first: monthly versus annual, regional pricing, currency, tax, a temporary promotion.
Then an enterprise plan, a product version, legacy docs, a usage tier, a publication date, a disclosed exception. What survives all of that becomes a finding. Nothing else does.
04 the refusal
A tool that can say no.
Anyone can generate a plausible-sounding audit. The hard part is a report where every sentence can be clicked and checked, and where "nothing found" is an allowed answer.
Findings below 90 confidence never reach you. When your pages agree, the report says exactly that, and lists what it read.
No high-confidence contradiction was verified during this scan.
34 public pages read · 18 claim types compared · 2026-08-24 14:02 UTC
Your public pages agreed with each other on 2026-08-24. That is rarer than it sounds.
Anonymized sample of a clear result.
No score. No grade. No gauge, no red dial, no invented figure of what this costs you. We show findings, or we show nothing.
05 how a scan works
One input. No account, no install, no repository access.
01
You paste your domain.
That is the only configuration. No page selection, no crawl depth, no claim-type checkboxes. The defaults are the point.
02
We read your public pages.
Public pages only, robots.txt respected. Nothing behind a login, nothing bypassed. You watch the pages being read.
03
You get findings with evidence.
Each one re-checked twice against your live pages, with the contradiction type, the confidence, and text ready to publish.
A free check takes a few minutes and reads up to ten public pages. You watch each page being read, and the result arrives by email as well.
06 the products
Three things to buy, once. No subscription.
Free Consistency Check
€0
One scan per domain. At most one verified finding, shown in full with its evidence and its proposed correction.
Run a free checkTruth Audit
€149
Every verified finding with dated evidence, outdated and ambiguous information, 20 simulated buyer questions, corrected text, executive summary. Refunded in full if fewer than three verified findings.
Buy Truth Audit · €149Truth Monitor 90
€499
The full audit, then rescans roughly every 30 days for 90 days: what changed, new findings, before-and-after evidence, and a final 90-day report.
Buy Truth Monitor 90 · €499One-time payment, no subscription. A renewal always needs your approval; we never charge automatically. Payment is taken by NanoCorp on a hosted checkout page, so no card details reach this site.
nothing to join
no card, no account
Run it on your own domain.
A domain and a work email. If your pages agree, the result says so and you have lost four minutes.
Or keep the link and send it to the founder whose pricing page you have doubts about.