RECORD / ENTRY 25
How an AI assistant turns one customer question into many smaller searches, and where each of the five AEO signals enters the answer it writes.
Bay Area AEO explains query fan-out in a sentence on its home page: AI assistants don't run one search. They break a question into many smaller searches, check different sources for each, then combine the results. The agency lays the process out in four steps: understand, fan out, retrieve, answer. This entry follows one Bay Area question through all four.
Take the question from the agency's dentists page: "Can you recommend a dentist in San Mateo who's good with anxious patients and takes Delta Dental?" The assistant first works out what is being asked. There is a service (a general dentist), a place (San Mateo), a quality (good with anxious patients) and a constraint (accepts Delta Dental). Each part has to be satisfied for a name to make the answer.
The assistant then splits the question into smaller searches, one or more for each part. It might look for dentists in San Mateo, for San Mateo practices that list Delta Dental, for reviews that mention nervous or anxious patients, and for local guides to dentists on the Peninsula. Each search can reach different sources, which is why one question can draw on a dozen pages.
| Step | What the assistant does | Signal that helps |
|---|---|---|
| Understand | Identifies service, place, quality and constraint | Answer-ready website that uses the customer's words |
| Fan out | Runs smaller searches for each part | Structured data stating services, area and hours |
| Retrieve | Reads listings, reviews and outside pages | Consistent listings; trusted mentions |
| Answer | Combines what it found and names two or three businesses | Specific reviews it can quote as a reason |
For each smaller search the assistant retrieves sources: business listings on Google, Bing, Apple Maps and Yelp; the practice's own website; directory profiles such as Healthgrades and Zocdoc; local press and guides. A practice whose listings agree on its name and hours, whose site states that it takes Delta Dental, and whose reviews mention gentle care with anxious patients turns up in several of these searches at once.
Finally the assistant combines the results and writes a short reply, usually with two or three names and a reason for each. A business that appeared across many of the smaller searches, with the same facts each time, is the easiest for it to recommend with confidence. The reason it gives is often drawn from review language, which is why the agency's review process invites customers to describe what the business does best in their own words.
Every question in a prompt panel goes through these steps. "Emergency plumber in San Jose open on Sunday" splits into searches for plumbers in San Jose, for emergency service and for Sunday hours. "Dog-friendly wineries in the Russian River Valley" splits into wineries in the valley and their pet policies. In each case, the business that states the condition plainly, on its site and its listings, appears in more of the smaller searches.
Fan-out explains why AEO is several jobs and not one. A business has to be findable in each of the smaller searches, and each search reads different sources. Listings help in one, a service page in another, a review in a third, a neighborhood guide in a fourth. Bay Area AEO's four services map onto this: the audit finds which searches miss the business, the cleanup and content fill the gaps, and the mentions and reviews give the final answer its reason. The monthly report's cited-sources field shows which of the smaller searches the assistant leaned on. The reference site covers the concept in its query fan-out entry, and its entry on how AI assistants choose businesses takes the same idea further.