45%
of US consumers used ChatGPT or another AI tool for local business recommendations in the past year, up from 6% the year before.
BrightLocal, Feb 2026, n=1,002
97%
of the businesses an assistant named in our 32-run probe across twelve verticals were picked from their own websites, not a marketplace.
Picked by Agents probe, Sept 2026
4%
of local queries get the same answer from every AI engine. On 48% they disagree completely. The pick is open.
Yext, 491,972 local queries, Feb to Aug 2026
What an answer engine actually does with your business
A customer asks the assistant a question with a place and a need in it. The assistant searches, opens a handful of sites, reads your Google Business Profile and a directory or two, and writes one answer with a few names and a reason for each. In BrightLocal's study of 200,000 local searches, 93% of the domains it cited were business websites.
So the answer is built from what you publish. Most answer engine optimization treats that as a content problem and sells articles. In our runs the assistant is not looking for articles. It is looking for a price, an opening, a service area, a phone number that matches, and a way to act.
“The fact the customer most needs is usually locked behind a call or a form. The business that publishes it gets picked.”
How we optimize for the answer
- 01
Start from the real question
Not a keyword. The question a customer types, with their ZIP code and their problem, run through the assistants people use in your market and against the businesses that beat you.
- 02
Read the assistant's reasons
Every run records what the assistant opened, what it passed on and why, in its own words. That is the fix list.
- 03
Publish the deciding fact
Price, availability, service area, what is included. On your own pages, in plain text, where an assistant finds it on the first pass.
- 04
Make the profile agree with the site
Google Business Profile, the directories the assistants cite, hours, phone, categories. One version of the truth everywhere the assistant looks.
- 05
Give the assistant a way to act
A booking or quote path an agent can complete without a phone tree. On Picked we rebuild it end to end and tag AI-sourced inquiries in your call tracking.
- 06
Run it again next month
The assistants change weekly. We re-run the same questions monthly, report who got named, and fix what moved.
Why we measure picks, not mentions
AI visibility tools count how often your name appears in answers to prompts a marketer wrote. A mention in a listicle answer is not a customer. A pick in answer to a real buying question is.
We report three numbers, always separately. The AI referrals we observed. The customers who told you an assistant sent them. And our estimate of new business. Never one inflated figure.
Typical AEO vs Picked by Agents
| Typical AEO | Picked by Agents | |
|---|---|---|
| Unit of work | Articles, backlinks, brand mentions per month | The customer question, and the fix that wins it |
| What is measured | Share of voice across prompts | Whether the assistant named you, and why, run repeatedly |
| Where the work lands | Blog posts and third party sites | Your service pages, your Google profile, your booking path |
| Guarantee | Traffic growth | Named within 90 days or we work free |
| Price | $1,597 to $4,000 a month per site | $2,997 or $4,997 a month, month to month |
Questions we get asked
- What is answer engine optimization?
- Answer engine optimization, or AEO, is making a business the answer an AI assistant gives when someone asks it a question. Unlike search engine optimization, there is no page of results to rank on. The assistant reads your site and profile and names a few businesses. AEO is about being one of them.
- How is this different from AI SEO or generative engine optimization?
- The terms overlap. AI SEO usually means using AI tools to do SEO. Generative engine optimization means getting cited by generative answers. We use answer engine optimization to mean one thing, being named when a customer asks, and we measure exactly that.
- Which assistants do you optimize for?
- The ones customers use. ChatGPT reaches 44% of US adults, Google's AI answers are read by 60%, then Gemini. We also run Meta Muse, Grok and Instinct, which act on the answer and book. The fixes are the same, because every assistant reads the same site and profile.
- How long until it shows?
- Assistants that browse live reflect a site change within days. Google profile changes take longer. We agree a baseline before we start and re-run the questions monthly. The guarantee is named within 90 days or we work free until you are.
- Do you write content?
- We write the pages the assistant needs to name you, such as a pricing page, a service area page or a booking path. We do not sell articles by the dozen. If your agency writes content, keep them. We make sure what they write carries the facts that win the pick.
Find out what the answer engines say about you.
The report is free. It shows who the assistants named for your customers' questions and where they stopped on you. Then we fix it, guaranteed.
Or write to Greg at greg@gregmiller.ai.