How to show up in ChatGPT and AI search as a small business

How does a small business show up in ChatGPT and AI search?

You show up by being findable, consistent, and quotable across the open web. AI assistants build answers out of pages, profiles, directories, and reviews that already exist, so there is no submission form and no paid placement to buy. The work is letting the crawlers read you, answering real questions in plain words, and making every detail about your business agree everywhere it appears.

That is the whole thesis, and I want it said in the first paragraph because it is the part most owners never hear. The market around this topic has filled up fast with tools promising AI visibility as a product. Almost all of them are selling a dashboard on top of work you can understand in an afternoon.

What has genuinely changed is customer behaviour. People used to type three words into a search box. Now a fair number of them type a full sentence into an assistant: I need someone to fix a leaking hot water tank in Castlegar this week, who is good. The assistant answers with a handful of names and a few links. If your business is legible to that process, you get named. If it is invisible, you get skipped, and you never see the missed call to know it happened.

Nobody submits a business to ChatGPT. The assistant reads the same web your customers do, so the job is to be worth reading and easy to quote.

How do AI engines pick which businesses to mention?

Most assistants answer a business question by running a search behind the scenes, pulling a handful of pages, and writing a summary from what they find. Your job sits in two places: being in the set of pages retrieved, and being the page whose wording is clean enough to lift. Miss the first and nothing else matters.

Think of it in two moves. The retrieval move decides which pages get opened at all, and it looks a lot like ordinary search: relevance, reputation, and whether the page can be crawled. The generation move decides which of those pages gets quoted and credited, and it rewards pages that state an answer plainly, in one place, without asking the reader to assemble it from three paragraphs.

This is why a business can rank respectably on Google and still never get mentioned by an assistant. The retrieval move found you. The generation move looked at a page of marketing adjectives, found nothing quotable, and used a competitor who happened to write a clear sentence about pricing. The full comparison lives in Google ranking versus AI answers, because the two really are separate fights with a shared foundation.

What does Google actually document about optimising for AI answers?

Google says there is nothing special to do. Its own guidance on AI features states there are no additional requirements to appear in AI Overviews or AI Mode and no special optimizations necessary, because the generative features run on the core Search ranking systems. Ordinary fundamentals plus a complete Business Profile is the whole strategy Google documents.

I lead with that because it saves people money. No special schema type unlocks AI Overviews. No chunking your content into tiny fragments. No buying mentions on content farms, which Google specifically calls less helpful than it might seem. The advice industry moved faster than the evidence here, and a lot of what gets sold as AI optimisation is regular search work with a new invoice line.

Where it gets honest is that Google speaks only for Google. ChatGPT, Claude, and Perplexity each run their own retrieval, and none of them publish a ranking guide. That gap is exactly where confident-sounding claims breed. My rule for this guide, and for what I tell clients: if a claim has no documented source or measured study behind it, it gets labelled as opinion or it does not get said.

Can AI crawlers read your site at all?

This is the first thing to check and the one that silently disqualifies businesses. If your robots.txt blocks the AI crawlers, or your pages only assemble their content after a script runs, you are invisible to the engines no matter how good the writing is. Access first, then quality.

GPTBot and OAI-SearchBot

OpenAI runs several crawlers. GPTBot gathers content, and OAI-SearchBot is the one tied to search results inside ChatGPT. It is the agent most robots.txt files forget, and forgetting it is the most common self-inflicted wound in this whole topic.

ClaudeBot and Claude-User

Anthropic documents its crawlers publicly, including which one fetches a page because a person asked about it in a conversation. Both read robots.txt.

PerplexityBot

Perplexity crawls to build the index it answers from, and it shows its sources beside the answer, which makes it the easiest engine to check yourself.

Google-Extended

This one gets misread constantly. It governs Gemini training and grounding. Google states plainly that it does not affect inclusion in Google Search and is not a ranking signal, so blocking it does nothing to your AI Overviews.

There is one trap here worth more than the rest combined. The moment you create a named group for a crawler in robots.txt, that crawler ignores your catch-all group entirely and reads only its own rules. So a robots.txt with a friendly catch-all and one hastily written named group can lock out the exact crawler you meant to welcome. Read the whole file, not the line you just added.

The second access question is whether your content exists in the HTML that arrives on the first request. Plenty of modern sites build their text in the browser afterwards. Google will usually get there eventually, since it queues pages for rendering, and the other engines are far less patient. View source on your most important page and search for your key sentence. If it is missing, that is your project.

On my own builds this is settled before launch: pages ship as static HTML with the answer in the source, and my robots.txt allows GPTBot, PerplexityBot, ClaudeBot, and Google-Extended by name, with a catch-all that welcomes everything else. If yours has never been looked at, what actually happens when someone Googles your business walks the same territory from the customer side.

How do you write a page an AI can quote?

Write the way a good employee answers a customer at the counter: name the question, answer it completely in the first breath, then add the detail. A model lifts self-contained passages. A paragraph that only makes sense after the two above it gets skipped, no matter how well written it is.

1. 01

   Answer in the first forty words

   Open every page and every section with a paragraph that answers the heading completely, without needing the paragraph before it. That block is the piece a model lifts. If it depends on context above it, it gets dropped or mangled.

2. 02

   Write the heading as the question

   People ask AI in full sentences. Headings phrased the way a customer would say it out loud give the engine an obvious match, and they help the human skimming on a phone at the same time.

3. 03

   Put comparisons in a table

   Versus questions are a huge share of what people ask an assistant. A clean two-column table with a row per dimension is the easiest structure in the world to extract, summarise, and attribute.

4. 04

   Date the facts that move

   Prices, rules, hours, and platform behaviour all drift. A stated date turns a claim into a checkable fact, and checkable facts survive summarisation with the source attached.

The lovely part of this is that the same edits raise your conversion rate for humans. People skim on phones. An opening line that answers the heading serves the reader in a hurry and the machine summarising the page, and both leave with the same impression of a business that knows its own answers. A proper questions and answers block does double duty here, which I unpack in what a great FAQ section actually does.

One warning about markup. FAQ structured data no longer produces a rich result in Google, so it is not the win it was sold as for years. Keep the visible questions and answers, because customers read them and models quote them. Treat the JSON-LD as tidy housekeeping.

What matters off your own website?

A surprising share of this fight is decided on pages you do not own. Assistants cite where you appear as much as what you host, so your Google Business Profile, your reviews, your directory listings, and your local press mentions all feed the answer. Consistency across those sources is the signal that makes an engine confident enough to name you.

• A complete Google Business Profile with the right categories, real hours, and a service area set by town, since Google names categories as the one profile field it ties to local ranking.
• The same business name, address, and phone number everywhere you appear, so nothing an engine reads contradicts anything else it reads.
• Real reviews from real customers, on the platforms your customers already use.
• Listings in the directories that genuinely serve your area, chamber pages, tourism sites, trade associations, and supplier locators.
• A presence on the one or two social platforms you will actually maintain, pointing back to the site you own.
• Mentions in local press and community pages, earned the slow way, because engines cite where you appear and not only what you host.

Categories deserve a special mention because Google is unusually direct about them. The categories on your Business Profile affect local ranking, and the test is whether your business is a thing, never whether it has one. A cafe that also sells firewood is a cafe. Google infers upward on its own, and a pile of loosely related categories dilutes the one that matters. My Google Business Profile guide for Kootenay businesses covers the whole setup, and it is the single highest-return hour in this entire guide for most local operators.

Contradiction is what quietly kills you. Old hours on a directory, a phone number from two owners ago, a price on a third-party listing that has not been right since 2023. An engine reading three versions of your business has to pick one, and the safest thing it can do is name someone else.

How do you tell whether any of this is working?

Manually, and monthly. Search Console folds AI Overviews traffic into the regular Web search type, so there is no separate report to open. The honest method is to ask the assistants the questions your customers ask, write down who gets named, and watch that list change over the season.

1. 1Write down the five questions a customer would genuinely ask an assistant before hiring someone like you, in their words, with your town in them.
2. 2Ask all five in ChatGPT, Perplexity, and Google AI Mode on the same day, and screenshot the answers.
3. 3Record which businesses were named, which pages were cited, and whether you appeared at all.
4. 4Read the pages that beat you. Nine times out of ten they answered the question in the first paragraph and you buried it.
5. 5Fix one page, wait a month, and run the same five questions again.

Two numbers get quoted constantly in this space and both deserve their source attached. One 2026 analysis published by SEO.com found the overlap between Google top ten results and what AI engines cite has fallen below 20 percent, and a related write-up from Goodfellas Tech reports that roughly 90 percent of ChatGPT citations come from pages outside Google top 20. Both are published by companies that sell AI search services, so hold both of them loosely. The reason I quote them at all is that they match what you see running the test above yourself.

On local questions specifically, I will say the thing most of this industry will not: nobody credibly knows yet. The one solid study is Pew, which tracked 900 US adults across 68,879 searches and found link clicks fell from 15 percent to 8 percent when an AI summary appeared. No Canadian sample, no local-intent breakout. A search for a plumber in Trail still lands in the map pack, where reviews and your profile decide it.

Where should a small business start?

Start with access, then structure, then consistency. Confirm the crawlers can read you, rewrite the opening paragraph of your three most important pages so each answers its own question, then hunt down every place your details disagree. That sequence costs almost nothing and covers most of the ground.

1. 1Ask three AI assistants the question your best customer would ask, in your town, and write down every business they name and every page they cite.
2. 2Open your robots.txt and confirm the AI crawlers are allowed, remembering that the moment you name a crawler in its own group, that crawler stops reading your catch-all rules entirely.
3. 3View the raw source of your two most important pages and check the answer is in the HTML that arrives, before any script runs.
4. 4Rewrite the opening paragraph of those pages so the first forty words answer the page question standalone.
5. 5Fix every place your hours, service area, phone number, or prices disagree across your site, your Google Business Profile, and your directory listings.
6. 6Add a visible questions and answers block using the exact wording customers use, with every answer written out on the page itself.
7. 7Repeat the assistant check monthly and keep the notes, because that record is the only measurement you are going to get.

Notice how much of that list is ordinary craft. Clear writing, honest details, a site that loads as real HTML, a Business Profile that is actually filled in. The engines changed, and the work that earns attention did not. If you would rather see the broader picture of how a site earns its keep, the difference between a website and a growth system puts this piece in context.

If you would like it handled, here is the honest shape of working with me. Already have a site? The AI and systems assessment is $300, credited in full against a build, and it maps where your visibility, structure, and details are leaking before anyone touches code. No website yet? I build custom sites from $2,000 for a Trailhead and $6,500 for an Engine, and every one of them ships with static HTML, answer-first pages, an open robots.txt, and the machine-readable files described in my guide to llms.txt. Fill out the contact form and tell me what you sell and who you sell it to, and I will tell you which of those two is the right move.

Sources and further reading

• Google Search Central: AI features and your website

  Google states there are no additional requirements or special optimizations needed to appear in AI Overviews and AI Mode. The generative features run on the core Search ranking systems.

• OpenAI: overview of OpenAI crawlers

  The official list of OpenAI user agents, including OAI-SearchBot, which is the one connected to search results inside ChatGPT.

• Anthropic: does Anthropic crawl the web, and how can site owners control it

  Anthropic documents its crawler names and how robots.txt controls them.

• Google Business Profile: improve your local ranking on Google

  Google names relevance, distance, and popularity, and says the categories you select affect local ranking. Local questions still resolve in the map pack for most searches.

• Pew Research Center: users are less likely to click links when an AI summary appears

  Browser-tracked study of 900 US adults across 68,879 searches. Link clicks fell from 15 percent to 8 percent when an AI summary appeared. No Canadian sample and no local-intent breakout, so read it as direction and not as your number.

• SEO.com: ChatGPT search statistics

  The source of the overlap figure quoted in this guide. Worth reading and worth remembering that it is published by a company selling AI search tooling.

Frequently asked questions