How to get your business mentioned by AI, fastest moves first.

There is no form that submits your business to ChatGPT, and no ad budget that buys a mention. AI engines name businesses they can verify (structured data and consistent facts), retrieve (pages in the indexes they read, shaped to be quoted), and corroborate (reviews, directories, and communities that talk about you). All three are buildable. This page orders that work by time-to-impact — what to do this week, this month, and this quarter — with an honest read on how fast each layer tends to pay off.

If you're reading this, you've probably already run the test: asked ChatGPT who's best in your category and read a list of competitors — or nothing at all. You're not late, but the clock is real: 45% of US consumers used AI to find local business recommendations in the past year, up from 6% the year before (BrightLocal Local Consumer Review Survey, 2026 — the shift behind the six-reason diagnostic).

What follows is the same list an AEO agency works through on a $2,000–6,000-a-month retainer (2026 market research), sequenced for an owner running it solo. If you want the deeper practitioner version of every step — scoring, diagnosis, the 25-point checklist — that lives in the complete AEO playbook. This page is the speed order.

Kill the myth first: you can't submit your business to ChatGPT

Search this topic and you'll find pages implying a “submission” process. There isn't one. Unlike Google Search Console or a business directory, ChatGPT, Gemini, and Perplexity have no portal where a business registers itself — and, as of mid-2026, no paid placement that buys an organic recommendation either (a point PR Newswire's own resource page on the topic makes plainly).

The most common piece of bad advice is “create a Custom GPT.” A Custom GPT is a private chatbot you configure inside your own account. It changes nothing about what ChatGPT tells everyone else who asks about your market. The same goes for telling ChatGPT to “remember” your business in a chat — that conversation ends where it started.

What functions like submission is becoming part of the material engines already read. Three currencies: verifiable facts (schema and consistent name-address-category data), retrievable pages (in the indexes engines search, shaped so one passage answers one question), and third-party corroboration (reviews, directories, Reddit threads, roundup lists). The mechanics of how engines weigh those inputs get their own page — how ChatGPT actually picks businesses — but you don't need the theory to start. Everything below maps to one of the three.

This week: make your business verifiable

The first horizon is entirely under your control — no waiting on crawlers, reviewers, or editors. Three fixes, roughly an afternoon each.

1. Ship schema — with the correct subtype

Structured data is the machine-readable layer under your pages, and generic tags waste it. A dental practice should carry Dentist schema, an HVAC company HVACBusiness— the specific LocalBusiness subtype for your category — with the same name, address, phone, and hours the visible page shows. Engines cross-check the markup against the page; schema that claims what the page doesn't say is worse than none. Service businesses should add a schema block per service, and any page built around questions should carry FAQPage markup.

Size the claim honestly while you do it. Google states that no special markup is required to appear in AI Overviews or AI Mode, and nobody has published an “AEO schema” that engines reward. What markup does is state your prices, hours, services, and Q&A in a form that leaves a parser nothing to interpret — and it still earns classic rich results in ordinary search, which is a return you can see without believing anything about AI at all. That is why it sits in week one: cheap, verifiable, and useful whether or not the AI upside materializes.

2. Align your Google Business Profile with the question you want to win

Your GBP feeds the entity data AI systems lean on for “near me” and local-intent questions: primary category, services, hours, and the exact business name, repeated consistently everywhere your business appears. Set the primary category to match your money question — an “HVAC contractor,” not a “contractor.” How the local answer stack fits together (GBP, reviews, NAP, schema) is covered in how AI picks local businesses.

3. Let the search crawlers in — they are not the training ones

Open yourdomain.com/robots.txt and read it with the distinction in mind, because the two kinds of crawler get confused constantly. OAI-SearchBot is what fetches pages for ChatGPT Search; Claude-SearchBot does the same for Claude, and PerplexityBot for Perplexity. Those are the ones a business wants open. GPTBot and ClaudeBot are primarily training-data collection, and blocking them does not remove you from the corresponding search surface — so if you have reservations about training, you can act on them without paying a visibility cost. Google-Extended is the same kind of training-side control.

The failure to look for: a blanket “block AI” rule, a CDN bot-protection default, or a security plugin that swept up the search crawlers along with the training ones. A crawler that can't fetch your pages can't quote them, and nothing on your site looks any different to you. Two minutes, and it occasionally explains everything.

This month: get retrievable — indexing first, then answer pages

The second horizon targets the live-retrieval pipeline — the one that moves in weeks rather than model-release cycles.

Get indexed where the answers get sourced — starting with Bing

Most businesses have ignored Bing through a decade of Google-only SEO. Two separate reasons to stop, and they carry different weights:

The second reason needs no interpretation at all: Microsoft Copilot answers from Bing's answer layer, Microsoft's own product on Microsoft's own index. So one effort buys a documented payoff on Copilot and a well-measured probability on ChatGPT — which is a better trade than most month-one work offers.

The fix is unglamorous: verify your site in Bing Webmaster Tools, submit your sitemap, and submit new URLs through IndexNow — which queues them within minutes, though what gets indexed is always Bing's call. The full walkthrough is in the Bing and Copilot guide, and if you would rather the submissions happened on their own, that is part of what our AEO service runs for you.

Publish answer pages — one question, one page

Retrieval-side engines quote passages, not domains. The page shape that gets quoted: the buyer's question as the title, a direct 40–80-word answer in the first paragraph, specifics after, FAQPage markup underneath. One page per question your buyers actually ask — “how much does AC repair cost in Tampa,” “Invisalign vs braces for adults” — not one page listing everything you do. If you don't know which questions matter yet, start with the five-prompt check to see where you currently stand, and write toward the questions where competitors get named and you don't.

This quarter: earn the third-party record

Here's the uncomfortable arithmetic behind the slow lane:

Reviews, at a steady rate

Engines read reviews the way buyers do — recent, specific, text-rich ones carry the signal. Build a habit of asking every customer at the natural moment (job done, order arrived), and respond to what comes in. Never buy or fake reviews: review platforms police it, and a purged profile reads worse than a thin one.

Directories and consistent listings

Find the directories and industry sites engines cite for your vertical — ask the engines your buyers' questions and note which sources the answers cite — then make sure you're listed there, with name, address, and category matching your site and GBP exactly. Stale listings with an old address or a dead phone number actively work against you, because engines cross-reference.

Show up where your category gets discussed

Community threads are among the most-cited sources in AI answers, Reddit above all — and there's a compliant way to participate that doesn't end in a ban, covered in the Reddit citations playbook. Alongside communities, pitch the roundup lists and real publications engines quote for your category. This is the slowest layer and the most durable one: practitioners put consistent recommendation — being named in more runs than not — at a 6–12 month build in most categories (Cited.so, 2026).

Which move matters to which engine

“AI” is not one machine, and the levers above don't pull evenly. The tendencies, engine by engine:

• ChatGPT. Two levers: OAI-SearchBot allowed in and quotable answer pages sitting in the general indexes its citations overlap with, plus — on a much slower clock — your overall public-web footprint, which is what model knowledge follows.
• Microsoft Copilot.Answers from Bing's answer layer, so the same indexing work counts here directly and without any inference.
• Google AI Overviews.Google's own index: schema depth, freshness, and decision-stage content — and a different contest from the Map Pack, as the AI Overviews guide breaks down.
• Gemini.Also reads Google's index, with entity consistency (site, GBP, listings agreeing) doing heavy work.
• Perplexity. Cites sources on every answer and favors fresh, structured pages and community sources — the most transparent engine to study, per the Perplexity guide.

How to know it's working — without fooling yourself

One manual check proves almost nothing, because the ground moves under every answer:

So measure like the engines behave: re-ask the same set of buyer questions on a schedule — monthly by hand, daily if software does it — and log two dates for every page you ship: the date it went live and the date an engine first cited it. On the fast lane, the observed timelines are shorter than most owners expect:

If a page goes uncited for a month, treat it as a diagnosis, not a failure: usually the question was wrong, the answer wasn't quotable, or the page never made it into the index — each of which the earlier sections cover.

The publisher's note

This playbook is published by AEO Action, software that runs the same loop on rails — that's the lens we write from, and the mechanics above are true whoever runs them. What the product adds: a daily scan asks the live engines your buyers' questions; an orchestrator picks one highest-value move per day and prepares the finished artifact — schema, a FAQ page, a landing page, or a Reddit pack — into an Approve Queue, and nothing publishes without your approval. Approved pages go out directly to Webflow and Squarespace; on WordPress, our plugin publishes approved pages and schema on its own schedule — we never push into your site; everywhere else you get finished HTML ready to paste, and we verify it went live. New URLs are submitted to Bing through IndexNow (Bing decides what to index), and every approved landing page is then watched against its target question for up to 120 days — so you get the dated receipt this page just told you to keep by hand: shipped on date A, first cited on date B.