How the AI-visibility audit works
Cited checks whether AI assistants recommend your home-services business when a customer asks — and hands you the fix list to get there. This page is the whole method, in the open: the exact questions we run, the engines we sample, what you get back, and what the check can and can’t tell you. An audit you can’t inspect isn’t worth trusting.
Nothing here is a black box or a scraped shortcut. We use official AI APIs, show you every question asked, and never invent a result. If a claim on this page isn’t something you could verify yourself, it doesn’t belong here.
How a Cited audit runs
- We ask your customers’ real buying-intent questions. Not invented keywords — the actual questions a homeowner types, like "who are the best HVAC companies in my city?", "which plumber should I call when a pipe bursts?", and "recommend a trustworthy, well-reviewed roofer near me". The same question set runs for every business in a trade, so one shop’s result is comparable to another’s.
- We sample each answer engine through its official API. Perplexity (Sonar) and ChatGPT web search answer those questions the way your customer would see them, and we record which businesses each one names. Using the official APIs — not a scraper — is what keeps the method clean enough to publish.
- We check where you land in each answer. For every question and engine we mark you cited (named with a link back to your site), mentioned (named, no link), or absent — and we capture which competitors get named instead, so you see who’s winning the recommendation you want.
- We turn it into a prioritized fix list. The report leads with the fixes that move you up fastest — a complete Google Business Profile, recent reviews, clear service-area pages, consistent listings, on-site FAQ. The same work wins you Google and the local map pack, so it pays off even where AI referral is still small.
Which AI engines we check — and which we deliberately don’t
- ChatGPT web search — measured, via the official OpenAI API. We query ChatGPT’s web-search answers through OpenAI’s API — the same retrieval a customer gets when they ask it for a local recommendation. ChatGPT corroborates across sources it retrieves, so a complete, consistent web presence is what earns the mention.
- Perplexity — measured, via the official Sonar API. Perplexity is citation-first: it answers with linked sources, which makes it the clearest read on whether you’re cited with a link or only mentioned. We sample it through the official Sonar API.
- Google AI Overviews — deliberately excluded. Google publishes no API for AI Overviews, and scraping them violates Google’s terms (an area under active litigation). We won’t build the product on a method we’d have to hide, so we don’t claim to measure Overviews. Our per-engine guide explains how the same fixes still influence them.
- Gemini grounding — deliberately excluded. Gemini’s grounded-answer terms forbid caching, analyzing, or reselling its results — which is exactly what an audit would do. So we leave it out rather than break the terms. If we can’t measure something cleanly, we say so instead of faking it.
- Sampled, not exhaustive. AI answers are non-deterministic — ask the same question twice and the names can change with phrasing, location, and the day. We sample and report a directional signal, never a guaranteed ranking. That’s exactly why the deliverable is a fix list, not a single scoreboard number.
What your report tells you
- Presence score. The share of the real buying-intent questions where an assistant named your business at all — your at-a-glance AI visibility.
- Who gets recommended instead. The competitors the assistants name for your city and trade, ranked by how often they come up — your AI share of voice.
- Cited with a link, or just mentioned. Whether an answer points back to your own site (a linked citation) or only says your name — the difference between a lead and a missed one.
- Your prioritized fix list. The specific, plain-English fixes that will move you up fastest — the same work that wins you Google and your Google Business Profile.
What this audit is not
- Not a guarantee you’ll be recommended. A clean report means you’re well positioned, not that an assistant will always name you. Answers vary by wording, location, and timing — being recommendable is what you can control, and what we measure.
- Not a live traffic number. We measure whether you’re recommendable, not how many customers arrive from AI. AI-referral volume to local businesses is still small today and we won’t pretend otherwise — the near-term payoff is that the same fixes win Google now, with a head start on AI as its usage grows.
- Not scraped or reverse-engineered. We never scrape Google AI Overviews or perplexity.ai, and never read data a service’s terms forbid. Everything we report comes from official APIs and is something you could check yourself.
- Not a data-harvesting funnel. Running an audit takes your business name, city, trade, and email — only so we can send you the report. No account, no card, and we don’t resell your details.
Run your free audit
See whether AI assistants recommend your business — free, no account. We email one report.
The exact questions we’ll ask AI about you:
- “Who are the best HVAC companies in my city?”
- “Which plumber should I call when a pipe bursts?”
- “Recommend a trustworthy, well-reviewed roofing company near me.”
Frequently asked questions
Which AI assistants does Cited actually check?
Two: Perplexity (via the official Sonar API) and ChatGPT web search (via the official OpenAI API). We deliberately exclude Google AI Overviews and Gemini — measuring either would mean scraping or breaking a terms of service, and we won’t build on a method we’d have to hide.
Why don’t you check Google AI Overviews?
There’s no public API for AI Overviews, and scraping Google’s results violates its terms (an area under active litigation). Rather than fake a measurement we can’t take cleanly, we leave Overviews out and are upfront about it. The good news: the same fixes that get you named by ChatGPT and Perplexity — a complete Google Business Profile, reviews, clear pages — are what influence Overviews too, so the work isn’t wasted.
Are the results exact or guaranteed?
No. AI answers are sampled and non-deterministic — ask twice and the names can change. We report a directional signal and a prioritized fix list, never a guaranteed ranking. That honesty is the point: anyone selling you a fixed "AI rank" is overselling.
Do you ever make up any of the results?
Never. If an engine doesn’t name your business, we report it as absent. The scoring is deterministic — the same answers always produce the same report — and we show you the exact questions asked and which engines gave them, so nothing is invented from thin air.
How long does an audit take, and what do you do with my details?
It’s early access, so audits are run by hand and we email your report within a few business days — no account, no sales call. We keep only your business name, city, trade, and email, purely to send the report. We don’t resell your details.