← Back to Signal Check

Sample Read

What lands in your inbox.

Below is the summary page of a real Signal Check, reproduced with the client’s identity and the specific claims withheld. Every figure on it is the figure that was reported. The full report runs a good deal longer; this is the page that gets read first.

Regional managed IT services provider · forty-eight answers · four AI models · client withheld

The Read in one page

A company its buyers know, and the models mostly don’t.

21%

of category answers named the company

10 of 48 responses

5 of 8

questions the company had a credible claim to win included the company

judged before the results were collected

7

direct factual contradictions

incorrect scale, facilities and certifications

6 → 1

mentions in the strongest AI model compared with the weakest

same company, same market

The marking, and what it caught

Marked winnable, before any answer came back

8 questions

Where the company actually appeared

5 of 8

Three named gaps on ground this company had a credible claim to. Those three carry the first three items on the fix list.

What the models got wrong

7 claims contradicted the record. They fell into three kinds:

  • Scale
  • Facilities
  • Certifications

The claims themselves are withheld here because they identify the client. In your report each one prints in full, beside the answer that produced it, the model and version that said it, and whether it was confirmed, contradicted or held for your confirmation.

What else is in the report.

  1. 01

    The question set, and the marking

    Every question printed in full, each one marked winnable, ceiling or exploratory before collection, with the reasoning for the marking. You can disagree with a call — you just can’t be surprised by one.

  2. 02

    Discovery

    How often you are named when the question does not contain your name. This is the clearest measure of unprompted discovery visibility, and it is kept strictly separate from questions that hand the model your brand.

  3. 03

    The names

    Every competitor and alternative the models raised, how often, and how often first. Including the ones you have never heard of, which is usually the interesting part.

  4. 04

    Findability

    Which domains the models actually read to answer questions in your category, and where yours sits among them. Helps distinguish a findability problem from a positioning, association or evidence problem.

  5. 05

    Name resolution

    Whether the models resolve your name, your product names, your abbreviations and your common misspellings to you. An abbreviation that resolves to someone else is a quiet, fixable leak.

  6. 06

    Factual accuracy

    Every claim a model made about you, checked against a fact sheet you supply. Confirmed, contradicted, or held for your confirmation — and anything we get wrong is withdrawn in writing.

  7. 07

    Closing the gap

    Priorities ranked by what is both important and realistically fixable, each one anchored to a measured figure rather than a best practice.

  8. 08

    Method and scope

    Model versions, dates, call counts, failures logged as failures, and what this measurement does not claim. Including the range around the headline figure.

Every figure in the report prints as n of N, with N named, so you can check it against the questions that produced it.

And every claim in it is read against the run that produced it by a person who knows your category, before it ships.

Request a Read →