Signal CheckSee if AI recommends you. And why.
Buyers are already asking ChatGPT, Claude, Perplexity and Gemini who to hire, what to trust and how firms compare. Signal Check shows whether you appear in those answers, who appears instead, what AI gets wrong and what is worth fixing.
Hover a column to see the buyer question behind it.
A real read
What a real Signal Check found.
A recent Read for a regional company in a certification-heavy category found:
Client identity withheld. Results were collected across Claude, ChatGPT, Perplexity and Gemini.
Reading the field
What the result tells you.
A visibility problem is not one problem. Signal Check separates issues that require different responses.
Missing
You do not show up at all
When the questions your buyers ask keep returning competitors, or no company at all, the AI models are not connecting your brand to work you have a credible claim to. That can point to missing evidence, weak entity signals, insufficient third-party authority, or content that does not clearly establish the connection.
Overlooked
You show up, and they still pick someone else
Sometimes the models know who you are and still recommend someone else first. That is a different problem. It can point to positioning, authority or evidence rather than simple visibility. More publishing is not automatically the answer.
Method
Built to tell you why.
Signal Check is designed to tell you what is happening and why, not simply chart a visibility score over time.
Decide what you can reasonably expect to win.
Before the test begins, we identify the questions your organization has a credible claim to answer, based on what you actually provide. That judgment is made before the results are collected and is not changed afterward, so the report distinguishes meaningful gaps from questions you should not expect to win.
Start with the buyer, not your marketing.
The question set is written without using your website copy or existing search rankings as the starting point. Questions reflect how buyers actually explore a category: category, qualified, capability, problem-led, comparison and verification.
Run the same questions across four AI models.
Every question is run against Claude, ChatGPT, Perplexity and Gemini, with model versions recorded. We capture whether you are named, where you appear, which competitors are named, and where a model fails to return a usable answer.
Find the evidence behind the answer.
Where the AI models disclose sources, we trace the pages, profiles, publications and mentions contributing to the result. That is what turns an AI visibility problem into something your organization can address.
What we measure
More than a visibility score.
Being named is only part of the picture. Signal Check also tests whether AI models can identify your company correctly, and whether the information they give a buyer is accurate.
Do the models recommend you?
We measure where your organization appears across the buyer questions that matter, how often it appears, and where it ranks relative to competitors.
Can they identify you by the name your market actually uses?
Each important version of your name is tested on its own. A company may resolve perfectly under its full legal or brand name while the shorthand its market uses every day returns the wrong company, or nothing at all.
Are the facts right?
Factual claims are checked separately against information you have confirmed. We count direct contradictions, not differences in tone or interpretation. Wrong locations, wrong scale, wrong pricing, wrong certifications, wrong capabilities. Those are accuracy issues, not visibility issues.
No false precision.
We report ranges where the underlying result can vary, and record the model version behind every answer. Automated scoring is checked against a fixed validation process, and we report any scoring decisions withdrawn during review.
From a recent Read
The full name resolved on every AI model. The short form resolved on none.
Every sales conversation, every conference badge and every internal document used the short form. To all four AI models it meant nothing, or it meant a different company.
What a score leaves out
No software can tell you which questions were winnable.
Free checkers from the large SEO platforms will tell you whether your brand appeared in an AI answer, and several paid dashboards will track that number over time. Both measure the same thing, which is appearance. Neither can tell you whether a question was one you had a credible claim to win, because that judgment depends on knowing your category, your buyers and what you actually sell — and it has to be made before the answers come back, or it is only a rationalization of the result.
Marked winnable, before any answer came back
8 questions
Where the company actually appeared
5 of 8
The score on its own was 21%, ten of forty-eight answers, and on its own that number says nothing about what to do. Read against the marking it says something specific: on the eight questions this company had a credible claim to win, it appeared in five. The other three are the work list.
Audience size and AI visibility are not the same thing.
A brand with real reach, real press and genuine authority can still be missing from the answer a buyer gets. In one recent Read, the client was named in 80% of answers about its category while a competitor with a far larger audience appeared in 5%. Being well known and being retrievable are different measurements, and only one of them shapes the shortlist.
A percentage with nothing under it is not a finding.
Every number in a Read prints as n of N, with N named in words, so you can check it against the questions that produced it. A score you cannot audit is a number you have to take on trust, and trust is the thing the whole exercise is supposed to replace.
Someone who knows your category reads it before it ships.
A model can be factually right about your company and still wrong about what the fact means in your market. Catching that takes a person who knows the category, and that is the step that does not scale.
The practice
D.J. Smith
Principal, CiV Digital
D.J. Smith is a brand strategist who has spent more than two decades helping leaders navigate growth and change.
His work spans positioning, naming, digital strategy and organizational change across nonprofits, media, professional services, manufacturing and emerging ventures. At the center of that work is a simple idea: find what is true, give it form, and help the world see it clearly.
Signal Check grew from that same problem. AI models are now interpreting organizations for buyers before those buyers ever reach the website. D.J. developed Signal Check to see what those systems understand, what they miss and what evidence is shaping the answer.
He reviews every Read before it ships.
The engagement
Start with the Read.
A complete diagnostic of how AI models find, recommend and describe your organization. No retainer required.
Start here
The Read
$2,500 · one time
- —A buyer question set built specifically for your category
- —Results across Claude, ChatGPT, Perplexity and Gemini, with model versions recorded
- —Every relevant mention, omission and competitor identified
- —Source tracing wherever the AI models disclose supporting sources
- —Name-resolution and factual-accuracy testing
- —Priorities ranked by what is both important and realistically fixable
Fix what the Read finds
The Retainer
$4,000 / month
- —A monthly working session with your team
- —Implementation of priority content and entity improvements
- —Two published assets
- —A re-run of the relevant question set to measure what changed
Every report is read against the underlying research by a person who knows your category before it ships. Nothing goes out on the strength of an automated score alone.
Repositioning, messaging or broader brand strategy is a different engagement. That is where Catalyst fits →
The Portfolio Read
For PR firms and private equity.
See the opportunity across the portfolio.
The Portfolio Read runs Signal Check across five to ten companies using the same methodology, creating a comparable view of how each is found, recommended and represented by AI models.
For PR firms
See where authority can change the answer.
Signal Check shows which competitors are being recommended, which sources are influencing those answers, where your client lacks authority, and where incorrect or incomplete information is being repeated.
That gives your team evidence to sharpen earned strategy, prioritize third-party authority and identify client work tied to an observable market gap.
For private equity
See where the portfolio can gain ground.
Compare how portfolio companies are being recommended, overlooked or misrepresented. Identify factual errors, weak category associations and gaps in the evidence AI systems rely on.
Use the findings to see where brand, reputation, content or digital investment may support commercial performance, and where no intervention is needed.
One read. A portfolio view.
The Portfolio Read includes:
- —A full Signal Check for each company
- —Side-by-side results using the same methodology
- —Competitor and category visibility
- —Name-resolution and factual-accuracy checks
- —Source and authority analysis where AI models disclose supporting evidence
- —Priority actions for each company
- —A working session with your team before findings are shared more broadly
Your relationship stays yours.
You choose the companies and remain the lead with clients or portfolio leadership.
We run the research, document the methodology and stand behind the findings. We can brief your team privately, join a presentation when useful, or hand you the work to carry forward.
Signal Check stays attached to the measurement. The strategy and relationship stay with you.
Start a Portfolio Read →FAQ
Fair questions.
01Who is this for?+
Organizations whose buyers do their research before they ever make contact. If a real person could ask an AI model “who should I use for this” and get back a list of names, there is something here to measure — and that covers a wider range of categories than most people expect, because the question set is built for your market rather than pulled off a shelf.
Where it earns less. If your growth is entirely referral and relationship with no research step in front of it, or your category is new enough that nobody has settled on a word for it yet, a Read will mostly confirm that. We would rather tell you on a call than have you find it in the report.
02Isn't this an AI-visibility tool I could rent for less?+
Those tools track whether you are named and chart it over time. Three things separate this from that, and all three are in the report.
The questions are written blind. Most tools build their prompt set from your own analytics: your top organic keywords, expanded. That asks your site what your site already answers, and companies score well on their own homework. Ours are written from what a buyer is trying to do, before anyone has seen your marketing or your figures.
What is winnable is called before the run. Every question is marked winnable or out of reach in advance, from what you actually deliver, and the marking is never revised. That is what turns a low score into reachable ground rather than a grade.
The scorer is audited by something that is not a model. A model reads each answer and records whether you were named. Every one of those records then passes a fixed check written by hand, and we report how many calls that check withdrew.
Once you know what to fix, a monitoring tool is a reasonable way to watch it, and we will tell you which one.
03Is this just SEO with a new name?+
No. SEO optimizes for a ranked page of blue links. This is about retrieval and entity: whether the models can find, trust, and assemble an accurate answer about you. Different mechanism, different fixes, and often a different team inside your building than the one that owns keywords.
04The answers change every time. How is this reliable?+
They do vary, which is exactly why a one-off screenshot is worthless. We report the headline figure as a range rather than a false-precision single number, so you can see how much a different but equally fair set of questions would have moved it. Model versions are recorded on every answer, and a failed call is logged as a failure rather than counted as an absence.
05Which models, and how many questions?+
A set built to your category, across six types (category, qualified, capability, problem-led, comparison, verification), run against four AI models: Claude, ChatGPT, Perplexity, and Gemini, with versions recorded. That's dozens to well over a hundred answers in a single read, sized to your market rather than a fixed list.
06What do I actually walk away with?+
The grid, the trace behind it, every competitor the models name in your category, priorities ranked by what's actually fixable, and a 90-minute readout with your leadership. It is yours to act on, with us or without us.
07Am I signing up for the monthly retainer?+
No. The Read is a one-time engagement at $2,500, complete in itself, with no monthly commitment and nothing to cancel. If you decide to close the gaps with us, the full $2,500 is applied to your first month of the Retainer, so if you continue, the read effectively costs nothing. You're never locked in; the credit is simply there if you want it.
08We represent several companies. Is there a version for us?+
Yes. The Portfolio Read applies the same methodology across five to ten companies, creating a comparable view across a client roster or investment portfolio.
PR firms can identify authority and reputation opportunities across accounts. PE firms can see which portfolio companies are being recommended, overlooked or misrepresented, and where intervention may create value.
09Where does Catalyst fit?+
Sometimes the Read identifies a tactical visibility or evidence gap. Sometimes it reveals a positioning problem underneath it.
The Retainer addresses the tactical gaps. Catalyst addresses positioning, messaging and broader brand strategy. The Read helps determine which problem you actually have.
10What if the Read says we're fine?+
Then you've spent $2,500 to know that with confidence, and we will tell you plainly, with no retainer pitch. More often it surfaces two or three fixable things worth far more than the fee. Either way, you leave with the truth about where you stand.
Start
Find out where you stand.
The answers are already being generated for your category. See what they say before you decide what to spend on. Tell us your category and a competitor or two, and a senior partner replies within two business days.
Prefer email? [email protected]