How AI Assistants Decide Who to Recommend (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews)

How do AI assistants decide which businesses to recommend?

AI assistants decide by aggregating signals. When you ask ChatGPT or Gemini "who is the best HVAC contractor in Phoenix," the assistant does not pull from a sponsored list or a curated database. It assembles a short answer from whichever businesses have the clearest, most corroborated digital footprint for that question.

The process has three stages: the assistant identifies what category the question is about, retrieves candidate businesses from the sources it trusts for that category, then ranks them on how confidently it can describe each one. Businesses with cleaner signals get ranked higher and named in the answer. Businesses with weak or contradictory signals get dropped, even if they are real and qualified.

What signals matter most?

Six signals matter most, in roughly this order of impact:

1. Entity clarity: one consistent business name, category, address, phone, and description across every site that mentions you. Inconsistencies confuse the model and cost you the recommendation.
2. Structured data on your site: Organization, LocalBusiness, Service, Product, and FAQPage schema. These tell the model facts in a format it can lift directly into an answer.
3. Third-party citations: profiles on directories the assistant trusts for your category (Google Business Profile, Apple Business Connect, industry-specific directories, review platforms).
4. Reviews: volume, recency, and the variety of platforms. Recent reviews on multiple sources beat a higher count on a single site.
5. On-page Q&A content: pages that answer the questions buyers ask, with the answer in the first 1 to 2 sentences under each heading so the model can extract it cleanly.
6. Source authority: how trusted the sites that mention you are within the model's view of the topic. A mention on a respected industry publication is worth dozens of generic directory listings.

These are signals AI reads automatically. Things like reputation, age of business, and ad spend can feed them indirectly, but none of them is what the model is actually inspecting.

Do AI assistants use the same signals as Google search?

There is overlap, but the targets are different. SEO signals optimize a single page to rank for a query so a user clicks through. AEO signals (the signals AI assistants use) optimize how the model understands and quotes your business inside an answer the user never has to click out of.

Two practical consequences: first, you can rank well on Google and still be invisible in ChatGPT, because Google reads link authority and on-page relevance while ChatGPT also weighs entity consistency and structured data. Second, you can be cited by an AI assistant without ranking on Google at all, if a directory or industry list with strong corroboration mentions you.

How do ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews differ?

All five weigh similar signals, but they retrieve and rank them differently. The short version:

• ChatGPT: when web browsing is enabled, it queries the web and reads pages directly. It tends to rely on a mix of search results and its own crawl. Recent additions to a site usually take 1 to 4 weeks to show up.
• Claude: similar to ChatGPT on browsing, with a slight bias toward authoritative and longer-form sources. Tends to be cautious about naming specific local businesses without strong citations.
• Perplexity: queries the live web on every request and shows its sources inline. The fastest of the five to reflect new schema or new citations.
• Gemini: pulls from Google's index plus its own retrieval. New schema added to a site is often reflected within days because Google crawls actively.
• Google AI Overviews: the AI block that appears above blue links in Google search results. Uses Google's live index, so what wins here closely tracks what ranks well in Google plus what has strong entity signals.

The practical takeaway: optimize for the signals all five share (entity clarity, schema, citations, reviews), and you cover the whole set. Optimizing for one assistant is rarely worth the work.

What can I do today to improve the signals AI sees?

Three concrete steps that take less than an hour each and move the needle on every assistant:

1. Add Organization and LocalBusiness schema to your homepage. State your name, category, full address, phone, hours, and a one-sentence description. Use Schema.org's official types. This single change closes the biggest signal gap for most businesses.
2. Audit your top 5 directory listings (Google Business Profile, Apple Business Connect, Bing Places, plus the top two industry-specific directories for your category) for consistency. Same business name, same category, same address, same phone. Fix any drift.
3. Add a 4-to-6 question FAQ section to your homepage that answers the questions buyers actually ask. Use FAQPage schema so the model can lift the answers directly.

These are the same fixes that show up in audit reports for most small and mid-sized businesses, in the same order. They are the foundation. More advanced work (publishing comparison content, building citations on niche industry sites, structured Product or Service pages) builds on top of them.

How long does it take for changes to show up in AI answers?

It varies by assistant. Perplexity often reflects new schema within 24 to 72 hours because it queries the live web. Google AI Overviews tends to follow within days for sites Google crawls actively. ChatGPT and Claude typically take 1 to 4 weeks, sometimes longer for sites that are not crawled often.

The right way to know is to test before and after. Ask each AI the same buyer query, write down which businesses it names, make the changes, wait two weeks, ask again. That is the loop the AI Visibility Monitor runs every month so you can see how the answers shift over time.