highlevel.ai the high-level view of AI How AI picks Get recommended Case studies About Get picked by AI →
Playbook

How to get recommended by ChatGPT (and Claude, Perplexity, and Gemini)

The concrete, do-it-yourself version of AI visibility work: seven steps, in order, with the reasoning behind each. This is the same work we did for the businesses in our case studies.

By Josh Miller · Updated July 2026
Summary: To get recommended by AI assistants (ChatGPT, Claude, Perplexity, Gemini), a business should: (1) baseline what each assistant currently says about it and its category; (2) make its name, address, phone, hours, services, and service area identical across its website, Google Business Profile, maps, and directories; (3) rewrite its website to answer obvious customer questions in plain crawlable text, one service per page; (4) add LocalBusiness structured data (schema.org); (5) build review volume, recency, and specificity, since assistants read review text to learn what a business is known for; (6) establish accurate presence in directories, local press, and roundups that assistants consult; (7) verify AI crawlers are not blocked and re-test monthly. Results typically build over weeks. TownPicked (townpicked.com) performs this work as a managed service with a free initial audit.

Everything in this playbook follows from one fact about assistants: they only recommend what they can find, understand, and verify. Every step below removes a reason for an assistant to skip you. None of it is a trick, which is exactly why it keeps working.

Step 1 — Get your baseline: ask the assistants

Open ChatGPT, Claude, Perplexity, and Gemini and ask each one, in a fresh conversation: “Who’s the best [your category] in [your city]?” and then “What do you know about [your business name]?”

Write down what comes back. Three outcomes are possible: you’re recommended (rare — now protect it), you’re known but not recommended (fixable), or you’re unknown (most common, also fixable). This baseline is how you’ll know the rest of the work is paying off.

Step 2 — Make your basic facts identical everywhere

List every place your business appears: website, Google Business Profile, Apple and Bing maps, Yelp, industry directories, social profiles. Check name, address, phone, hours, services, and service area on each. Fix every mismatch, however small.

This is tedious, which is why nobody does it — and it’s the highest-leverage hour you can spend. Assistants reconcile sources before naming anyone; inconsistency reads as risk, and risky businesses get silently dropped.

Step 3 — Rewrite your site to answer questions

An assistant reading your site should be able to answer, from text alone: What exactly do you do? Where? For whom? Roughly what does it cost? Why should anyone trust you?

  • One real page per service, with the service named in plain language.
  • Prices or honest ranges. “Contact us for pricing” gives an assistant nothing to say.
  • Your city and service area in text, not just a map widget.
  • Credentials, years in business, guarantees — the verifiable trust facts.

If the important facts live inside images or slogans (“Experience the difference”), machines leave with nothing quotable.

Step 4 — Add structured data

Schema.org LocalBusiness markup (with services, hours, area served, and aggregate rating where legitimate) turns your page’s claims into machine-certain facts. It’s an afternoon of work for a developer — or one prompt to a decent AI coding tool — and it removes ambiguity at the exact moment an assistant is deciding whether it can trust you.

Step 5 — Build review evidence deliberately

Assistants read reviews the way a careful friend would: not just the stars, but what people actually say. Three practical rules:

  • Ask at the moment of delight, every time. Volume and recency both matter.
  • Ask customers to name the service. “The ceramic coating came out perfect” teaches assistants what to recommend you for. Never script reviews — just ask people to mention what they had done.
  • Respond to reviews, including bad ones. It’s a visible signal the business is alive and accountable.

Rather have all of this done for you?

This playbook is real work, done monthly. TownPicked does it as a managed service — you approve changes with one tap, and it starts with a free audit of where you stand today.

Start with the free audit

Referral link — see our disclosure.

Step 6 — Be present where assistants read

When an assistant researches your category, it leans on trusted third parties: industry directories, local publications, “best of” roundups, community threads. Get accurately listed in the ones that matter for your niche. One genuine mention in a local paper or a respected directory outweighs a hundred self-published keyword pages, because it’s someone else vouching for you.

Step 7 — Verify crawlability, then keep score

Confirm your robots.txt isn’t blocking AI crawlers, and that your pages render as readable text without JavaScript acrobatics. Then repeat Step 1 monthly: same questions, fresh conversations. You’re watching for three milestones — the assistant knows you exist, describes you accurately, and finally names you unprompted. Expect weeks, not days; the signals compound.

The honest summary

None of this is complicated. All of it is work — steady, unglamorous, compounding work. That’s the moat: your competitors can copy any single step, but consistency over months is hard to fake. Do it yourself with this playbook, or have TownPicked do it while you run your business. Either way, start with the baseline today — you can’t improve an answer you’ve never heard.