Twelve checks that decide whether a brand shows up in Google AI Overviews and in the answers ChatGPT, Perplexity and Gemini give. Written so a client can read it without a glossary.
Search is no longer a list of ten links. Google answers many queries with an AI Overview, and a growing share of buyers ask ChatGPT, Perplexity or Gemini before they open a browser tab. Those answers name a handful of sources. A brand is either one of them or it is absent.
Being cited is not the same as ranking. The engines read a site the way a fact-checker would: can they parse it, can they tell who is behind it, is there anything on it worth quoting, and does it agree with what the rest of the web says about you. Most sites fail quietly on two or three of those.
The fixes are mostly ordinary. Clear pages, named people, original numbers, consistent details, and a monthly habit of checking what the engines actually say. This checklist turns that into twelve yes-or-no questions.
Rendering, entity clarity, structured data, consistent details.
Checks 1 to 4 · TechnicalOriginal data, author pages, first-hand experience, freshness.
Checks 5 to 8 · EditorialTracked prompts, share of citations, a baseline, a review cadence.
Checks 9 to 12 · MeasurementChecks 1 to 4 · Technical
If an engine cannot parse the site or tell who is behind it, nothing else on this list matters yet. These four are technical, and usually quick.
Key page text is in the HTML the server sends, not added by JavaScript later, and robots.txt lets Googlebot, GPTBot, PerplexityBot and ClaudeBot in.
TestOpen View Source, not Inspect, and search for a sentence from the page. Then read the Disallow lines in robots.txt.
The first sentence of the home page and the About page says who you are, what you do, for whom and where. One company name, used everywhere.
TestShow the home page to a stranger for ten seconds, then ask what the company does. If they hesitate, so will the engines.
Organization schema on the site, Person schema for leaders and authors, Product or Service schema on offer pages, all matching the visible text.
TestRun the home page and one service page through validator.schema.org. Look for Organization and Person with name, logo and sameAs.
Name, address and phone match exactly on the site, Google Business Profile, LinkedIn and key directories, and those profiles are listed in sameAs.
TestSearch the brand name and open the first five results. Compare name, address and phone line by line. One mismatch is a fail.
Checks 5 to 8 · Editorial
Engines cite sources, not summaries of sources. These four decide whether there is anything on the site worth quoting, and whom to credit for it.
At least one page publishes something only you could: a survey, a benchmark, your own numbers, a price list, a method and what it produced.
TestName three facts a journalist could quote from your site. If the list is empty, the engines have nothing to cite either.
Every article has a named author. Every author has a page with a photo, role, credentials, a LinkedIn link and Person schema.
TestClick the byline on your three most-read articles. No byline, or a byline that goes nowhere, is a fail.
The content shows you did the thing: screenshots, the process, the numbers, what went wrong. "In our tests" rather than "experts say".
TestTake your best article and strike out every sentence a competitor could have written. What survives is your experience signal.
Key pages show a real "last updated" date and sit on a review schedule. Pages untouched for two years are refreshed, merged or removed.
TestSort pages by last-modified date in your CMS. Count the ones that still get traffic and have not changed since 2024.
Checks 9 to 12 · Measurement
What gets measured gets fixed. These four turn AI visibility into a number on the monthly report.
A list of 20 to 30 questions your buyers ask, run every month in ChatGPT, Perplexity, Gemini and Google AI Overviews, with the answers saved.
TestWrite ten prompts, run them in each engine, and note in a sheet which domains each answer cites.
For those prompts you know how often you are cited against your three closest competitors, and which of their pages earn the citations.
TestCount citations per domain in the check 9 sheet. Your share of the total is the number to watch.
The first full month is recorded and dated, so every later month is measured against a number rather than a memory.
TestCan you say, with a date, what your share was when you started? If not, this month is month one.
Someone owns it. Prompts are re-run monthly, the list is revised quarterly, and the results sit on the same report as rankings and traffic.
TestName the owner, the day of the month it runs and the report it lands in. Three answers is a pass.
Your score
Not yet visible. The engines cannot read or verify the brand. Start with group A: mostly technical, weeks not months.
Readable, not cited. The engines see you but have no reason to quote you. Group B is the work: data, authors, experience.
Ready. Make it a habit. Group C monthly, on the same report as rankings and traffic, and this checklist again next quarter.
Ads and traffic generation
Readiness gets a brand cited. Traffic still has to be earned, bought and converted.