How ChatGPT, Perplexity and Google AI Overviews choose which brands to name

KEY TAKEAWAYS

1. AI engines retrieve a short list of trusted sources per question, then name the brands those sources agree on.
2. The same handful of domains keeps appearing for a given question. Those are the cited sources, and the press pillar is built on them.
3. Consistency of description across sources matters as much as volume.
4. Engines verify the entity before naming it: schema, Wikidata, LinkedIn and Google Business Profile have to agree.

Ask ChatGPT, Perplexity and Google AI Overviews the same buyer question and you get three different answers that name roughly the same brands. That overlap is the thing to understand. The engines differ in interface and style, but they share a method. It is the method that digital authority measures.

Step one: retrieve, do not rank

A search engine ranks every page it knows about and shows you the top ten. An AI engine does something narrower. It turns your question into a few searches, pulls back a short list of pages, usually fewer than a dozen, and reads them.

Those pages are the entire world the answer is built from. If your brand is not in them, no amount of quality on your own site changes the outcome.

Which pages make the list is predictable. The engines favour:

• Publications with a track record in the category: trade press, business media, review sites.
• Pages that directly address the question, often "best X for Y" roundups and comparisons.
• Recent pages. Last year's roundup loses to this quarter's.
• Pages the engine has cited before and that users did not push back on.

Run the same question ten times and the same handful of domains keeps appearing. We call those the cited sources for a question, and the press pillar of the Digital Authority Score is built on them.

Step two: read and tally

Once the engine has its sources, it extracts the brands they mention, how they describe them and how confidently. A brand that shows up in six of eight sources with the same description gets named first. A brand that shows up once, described three different ways, gets named last or not at all.

This is why consistency matters as much as volume. If one source calls you "payroll software", another calls you "an HR platform" and your own site says "workforce management", the engine has three weak signals instead of one strong one.

Step three: verify the entity

Before naming a brand, modern engines check that it is a real, distinct thing. They look for the records that machines use to confirm identity:

• Organization schema on your site with a consistent name, logo, and sameAs links.
• A Wikidata item, and ideally a Wikipedia article.
• A LinkedIn company page and a Google Business Profile that match.
• The same name, description and URL everywhere.

Brands that fail these checks get blurred into competitors or dropped. This is the cheapest pillar to fix and the one most brands ignore. It is 15% of the score. The five-pillar checklist lists the eight checks.

What this means for earning a mention

Put the three steps together and the playbook writes itself.

1. Find the cited sources for your target questions. Not "the press" in general. The specific publications that come back when the engine retrieves for your question.
2. Get into them. A real article in a publication that is already in the engine's retrieval set is the single highest-leverage move. It changes press coverage and AI visibility at the same time.
3. Say the same thing everywhere. One category label. One description. Use it on your site, in your press, on LinkedIn, in your schema.
4. Close the entity gaps. Schema, Wikidata, LinkedIn, Google Business Profile. A day of work.
5. Measure across engines and repeat. Twenty questions, three engines, monthly.

A worked example

A payroll provider wants to be the answer to "best payroll software for restaurants?". The report shows the engines citing eight sources for that question. The brand appears in one. Its leading competitor appears in six.

The fix is not a blog post about restaurant payroll. The fix is coverage in four of the seven sources the brand is missing from, with the same positioning in each. Two months later the brand is named on two of three engines, second position. The score moved from 38 to 54, and most of the gain came from one pillar.

That is the pattern. Find the sources. Get in them. Stay consistent.

Questions

Do AI engines use my website?

Sometimes, but rarely as the deciding source. Your site tells the engine what you say about yourself. Independent publications tell it what others say. The second carries more weight when the engine picks who to recommend.

Why do answers change between runs?

The engines sample sources and generate text each time, so the exact wording and order move. The set of brands named is more stable than the order. That is why we ask 20 questions across three engines and count, rather than screenshot one answer.

Is this just SEO with a new name?

It overlaps. Search presence and organic authority together count for 30% of digital authority and Google AI Overviews draws on Google's index. But the biggest lever has moved from on-page optimisation to being cited by trusted third-party sources.

Does paid media help?

Not directly. Ads are not sources. Paid placements in editorial publications, when they are real articles that the publication indexes, do count because the engine reads them as coverage.

Sources

1. Google Search Central, AI features and your website developers.google.com
2. OpenAI, Overview of OpenAI crawlers platform.openai.com
3. Perplexity, PerplexityBot and crawler documentation docs.perplexity.ai
4. Google Search Central, Introduction to structured data markup developers.google.com
5. Wikidata, Introduction wikidata.org

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →