Skip to content

Accuracy Check

Is AI saying something wrong about you

During the analysis, claims in the answers are compared with your brand's own pages. A contradiction becomes a step on the roadmap; a check that did not run is never written as a clean result.

Get a quoteCheck technical readiness

Free technical readiness check · no sign-up · no card

A claim and your own page

Example

  • “Sample Brand A is open at weekends.”

    matchescounted

  • “Sample Brand A only operates in Istanbul.”

    contradictscounted

  • “Sample Brand A offers free shipping.”

    nothing to compare againstnot counted

  • The check did not run on this answer

    did not runretried

Example view · drawn with a sample brand and sample data, not a real measurement.

Answers are collected on 5 engines, each one reported on its own:

  • ChatGPT
  • Gemini
  • Perplexity
  • Claude
  • Google AI Overview

A claim and your own page

Example

  • “Sample Brand A is open at weekends.”

    matchescounted

  • “Sample Brand A only operates in Istanbul.”

    contradictscounted

  • “Sample Brand A offers free shipping.”

    nothing to compare againstnot counted

  • The check did not run on this answer

    did not runretried

Example view · drawn with a sample brand and sample data, not a real measurement.

01

Every claim, against your own page

What the answer says about your brand is compared with what your site says. When a contradiction is found, the two are written side by side and marked low, medium or high by how much harm it would do if a buyer acted on it.

  • Comparison is made only for claims your site has a counterpart for.
  • A mismatch is not always the engine's fault; sometimes your site is out of date.

Report summary

Example

Summary

Sample Brand A is named behind its competitors in comparison questions

  • Visibility in buyer questionsmeasured
  • Mentions in price questionsmeasured · no mention
  • Newly added audiencenot measured yet
Example view · drawn with a sample brand and sample data, not a real measurement.

02

'We found nothing wrong' means three different things

The check may not have run, there may have been nothing to compare against, or there really is nothing wrong. Only the third is good news, and only the third enters the rate.

  • A check that did not run is not counted and can be retried.
  • If there is nothing to compare against, the result reads 'not measurable'.

Roadmap

Example

Now

A guide page answering the comparison questions

Next

Update the brand page on a review site

Research

Which source is read in price questions

Watch

The competitor named with you in trust questions

Example view · drawn with a sample brand and sample data, not a real measurement.

03

A contradiction becomes a piece of work

Contradictions the check finds appear as a step on the analysis roadmap. The roadmap also states whether the check ran.

  • The fix is made at the source, not in the engine: an old page, a missing fact, a record confusing you with another company.
  • If the check did not run, the roadmap says so; an empty list is not a clean bill.

Technical readiness

Example
  • ChatGPT's search crawler can reach the sitepassed
  • Main content reads without JavaScriptneeds fixing
  • Structured data (JSON-LD) is presentpassed
  • The sitemap is declared in robots.txtcould not read
Example view · drawn with a sample brand and sample data, not a real measurement.

04

The brand whose site blocks the crawler carries the most risk

If your site blocks our crawler, there is nothing to compare against and the check cannot run. The brand most at risk of being described wrongly and the brand whose check comes back empty are often the same, so the two are never written the same way.

  • The free check shows which AI crawlers cannot reach your site.
  • Who our crawler is and why it sends requests is set out on its own page.

What the report covers

Example
  • Visibility in AI answersmeasured
  • Google ad footprintmeasured
  • Meta adsnot measured yet
  • Accuracy checknot in this report
Example view · drawn with a sample brand and sample data, not a real measurement.

05

Where it runs today, and where it does not

The accuracy check runs today inside the scan-based analysis, and its result goes to the analysis roadmap. Registered measurement reports do not carry this section yet; the report says so plainly as 'not included in this report'.

  • A section missing from a report is never read as 'nothing wrong'.
  • The check's result is kept with the analysis's own records.

How it works

From claim to correction

  1. We separate the claims

    Statements about your brand in the answers that could be true or false are set apart.

  2. We compare with your page

    Each claim is compared with its counterpart on your site; if there is none, that is written apart.

  3. You fix the source

    The contradiction goes onto the roadmap; the fix is made where the error came from.

Method promise

A check that did not run produces no reassurance

This module exists because of one sentence: 'we found nothing wrong about you' can be written the same way when the check never ran. We keep the three states apart and write only a genuinely measured clean result as good news.

Read the method

What we measure

  • What the answers say about your brand
  • What your site says on the same point
  • How serious a contradiction is for a buyer
  • Whether the check ran

What we do not measure

  • Value judgements and opinions
  • Claims your site has no counterpart for
  • Accuracy in registered measurement reports: not included yet

Other modules (9)

One measurement, answers to other questions

Frequently asked questions

Can we get wrong information corrected?

The engines have no correction button. When we find where the error comes from, the fix is made there: an old page, a missing fact, a record confusing you with another company.

What if our site blocks the crawler?

The check cannot run, and the result reads 'not measurable', not clean. The free check shows which crawler is blocked; often a fix to a robots.txt rule is enough.

Which claims are compared?

Those that can be true or false: for example a price, the scope of a service, a location or whether a product exists. Value judgements such as 'user-friendly' are not compared, because they cannot be.

Why is the accuracy check not in the report?

The check runs today inside the scan-based analysis and its result appears on the analysis roadmap. Registered measurement reports do not carry this section yet, and say so as 'not included in this report'.

Find out what is being said about you

The check can only run if your site can be read. The free check measures that; to scope an analysis together, ask for a quote.

Get a quoteCheck technical readiness

Free technical readiness check · no sign-up · no card