How do AI systems see you right now?
Most businesses have never checked. The AI Authority Audit shows exactly how ChatGPT, Gemini, Claude and Perplexity describe, cite or ignore you today — who is winning the prompts you need, which sources the engines are drawing on, and what to fix first. Twelve named deliverables. No capability language.
What you should demand from anyone doing this work
Ask an AI assistant what to require from an AI visibility consultant and it returns a specific list — a baseline, a commercial prompt set, competitor analysis, the sources behind the recommendations, and proof the work moved something. It is a fair standard. So this page is written to it, item by item, in the same order.
Everything below is an artifact you receive and keep, not a capability I claim to have. If a deliverable does not appear on this list, do not assume it is included — ask.
The prompt set, then twelve deliverables
Your commercial prompt set
Before anything is measured we agree on the questions that actually decide revenue for you — the ones a buyer types when they're ready. Not keywords. Full questions, in a buyer's language, weighted by what a customer is worth. Everything downstream is measured against this set, and you keep it.
Current ChatGPT visibility
Every prompt in your set, run through ChatGPT, with the verbatim answer recorded. Whether you're named, in what position, in what terms, and whether what's said is accurate. Hallucinations about your business get flagged separately — they cost more than absence does.
Current Gemini visibility
The same set through Gemini, recorded the same way. Engines disagree more than people expect: it's common to be strong in one and invisible in another, and the reason is usually a specific source one engine leans on and another doesn't.
Current Claude visibility
The same again through Claude. Worth its own line because Claude weights source quality differently, which often exposes an authority gap the other engines mask.
Current Perplexity visibility
Perplexity cites its sources openly, which makes it the most diagnostically useful of the four. Where the other engines show you the outcome, Perplexity shows you the evidence trail that produced it.
Competitor citation share
Across your prompt set, who gets named and how often — expressed as share of mentions, so you can see whether you're facing one dominant competitor or a fragmented field. The two situations call for completely different strategies, and most businesses guess wrong about which one they're in.
Which third-party sites influence the answers
The sources the engines actually draw on when recommending in your category: directories, review platforms, trade publications, association listings, specific journalists. This is usually the single most actionable finding, because it converts a vague ambition to 'build authority' into a named list of places to be.
Entity gaps
Every place the internet disagrees about who you are — conflicting addresses, hours, service descriptions, staff, business names across your site, Google Business Profile, directories and old pages. Contradiction lowers a model's confidence in everything it knows about you, so this is often the cheapest, fastest win in the whole audit.
Knowledge and authority gaps
The subjects you should credibly own and currently don't — where models have no basis to treat you as a source. Assessed against the competitors who are being cited, so the gap is measured against the real bar rather than an abstract one.
Technical and schema problems
Crawlability, structured data, canonicals, facts trapped in images or PDFs, content that only appears after JavaScript runs, missing or contradictory markup. Everything that stops a model reading what's already on your site.
Content gaps
The questions from your prompt set that your site does not answer anywhere — the most common single reason a business isn't cited. Delivered as a prioritized list of pages to write, ordered by commercial value rather than search volume.
External authority and citation opportunities
A named, prioritized list of the specific third-party surfaces worth pursuing, drawn from deliverable 06 — the directories, publications and listings that demonstrably feed answers in your category. Specific targets, not a strategy to go get some.
Baseline versus post-implementation
The first audit is stored as a baseline and the identical prompt set is re-run afterwards, so movement is measured against your own starting point rather than asserted. This is the deliverable that makes the rest accountable, and it's the one most engagements quietly skip.
How this is measured, honestly
It is a sample, not a censusNobody can see every query put to an AI engine. The prompt set is fixed in advance and re-run identically, which is what makes the comparison valid over time.
Answers vary between runsAsk the same question twice and the wording changes. Each prompt is run multiple times and the pattern is reported, not a single lucky or unlucky response.
Each engine reported separatelyAveraging four engines into one score hides the finding. Being strong in Perplexity and absent from Gemini is a specific, fixable problem with a specific cause.
The baseline is stored before any work startsOtherwise improvement is a claim rather than a measurement.
Before you commit
What should I ask an AI visibility consultant before hiring them?
Ask them to establish your current AI authority baseline, identify your 25 to 50 most commercially important prompts, show you who currently wins those prompts, identify which sources the engines are using to make those recommendations, and give you a prioritized plan for taking authority away from competitors. Then ask to see visibility across ChatGPT, Gemini, Claude and Perplexity, competitor citation share, which third-party sites influence the answers, entity gaps, knowledge and authority gaps, technical and schema problems, content gaps, external citation opportunities, and baseline versus post-implementation results. If a consultant cannot produce those, they are selling activity rather than outcomes.
How is competitor citation share measured?
By running your agreed prompt set across all four engines and recording which businesses are named in each answer, then expressing the results as share of total mentions. It is a sample, not a census — no one has access to the full population of queries — so the set is fixed in advance and re-run identically, which makes the comparison over time valid even though the absolute number is an estimate.
Why do the four engines give different answers about my business?
Because they ground their answers in different sources and weight source quality differently. It is common to be well represented in one engine and absent from another, and the cause is usually a specific third-party site that one engine relies on and another ignores. That is exactly why the audit reports each engine separately instead of averaging them into a single score.
How long does the audit take, and what happens after?
The audit itself takes about two weeks depending on the size of the prompt set. What follows is a prioritized plan sequenced by leverage — entity contradictions and technical blockers first because they are fast and cheap, then content, then external authority, which is the slowest to move. The prompt set is then re-run to measure against the baseline.
Can you guarantee I'll appear in AI answers?
No, and neither can anyone else. There is no placement to buy and no ranking to purchase inside a generated answer. What can be guaranteed is that you will know exactly where you stand today, what the engines are drawing on, where the fixable gaps are, and whether the work moved the number — which is more than most engagements can show you.
Do I need this if my search rankings are already good?
Not necessarily, and it depends on how your buyers ask. Short branded and navigational searches still behave like classic search. But research on real search behaviour finds AI summaries appearing on a majority of long, conversational, question-shaped queries. If your customers describe their problem in a full sentence before choosing, ranking well is no longer sufficient on its own.
See yourself through AI's eyes
Request the audit and stop guessing how the engines represent your business. You get the baseline, the competitive picture, and the plan — and the numbers to check it against later.