How to get your business recommended by ChatGPT

Why this is suddenly worth doing

The behavior change is simple and it is already well underway. Someone who would previously have searched for a category, opened four tabs, and compared them now asks ChatGPT which business they should use and gets two or three names. If you are not among those names, you were never in the consideration set. There is no page two.

This is a genuinely different competitive surface from search, and right now it is much less contested. Most businesses have done nothing about it, including most agencies. That window will close, but it is open.

Pillar 1: Let the crawlers in

This is the most basic step and the most commonly failed one. A significant share of sites block AI crawlers, sometimes deliberately, more often because a security plugin did it silently or a developer copied a robots.txt from somewhere else.

If GPTBot cannot fetch your pages, nothing else in this article matters. Check yourdomain.com/robots.txt and confirm these are not disallowed:

• GPTBot: OpenAI's training and retrieval crawler
• OAI-SearchBot: OpenAI's search index crawler
• ChatGPT-User: fetches pages live when a user's question triggers browsing
• PerplexityBot: Perplexity's crawler
• ClaudeBot and Claude-User: Anthropic's crawlers
• Google-Extended: controls use in Gemini and AI Overviews grounding
• Applebot-Extended: Apple Intelligence

A brief note on the tradeoff: blocking is a defensible decision for publishers whose revenue depends on the visit itself. For a local business trying to be found, blocking is self-harm. If a model cannot read your site, it recommends the competitor whose site it could read.

One more piece of housekeeping: many AI crawlers do not execute JavaScript. If your content only appears after client-side rendering, it may be invisible to them even when they are allowed in. Server-rendered HTML is the safe choice.

Pillar 2: Write so an answer can be lifted out

Models do not read appreciatively. They extract. Content that gets extracted has a recognizable shape, and it is not the shape most marketing copy takes.

Answer in the first hundred words

State the complete answer immediately, before the context, the background, or the story. If someone asks how long does local SEO take, the page should say typically three to six months for competitive terms, with first movement in 30 to 60 days near the top, not in paragraph nine after a history of Google's algorithm.

This is also better for humans, which is a useful property of most GEO advice.

Use headings shaped like questions

People ask AI models questions in full sentences. Headings that match those sentences give the model an obvious anchor. How much does a website cost in Renton? outperforms Pricing, because the first one matches the query and the second requires inference.

Make paragraphs self-contained

Here is a useful test: take any paragraph out of your page and read it alone. Does it still answer something? If it only makes sense given the three paragraphs above it, it will not be cited, because extraction pulls fragments rather than whole documents.

Say the causal thing out loud

Human readers infer. Models do better with explicit statements. Reviews improve Map Pack ranking because review signals are a confirmed local ranking input, and recency now carries more weight than total count is more extractable than reviews matter a lot, because it states the mechanism.

Mark the structure up

FAQPage and HowTo structured data label your content machine-readably. This is not a magic ranking lever, but it removes ambiguity about what is a question and what is its answer, and removing ambiguity is most of this job.

Pillar 3: Become an entity a model can resolve

For a model to recommend you confidently, it needs to be sure that the business named in a directory, the one mentioned in a forum thread, and the one on your website are all the same organization. Ambiguity here is expensive: two variants of your name across the web produce two half-strength entities instead of one strong one.

What builds a resolvable entity:

• Exact name, address, and phone consistency everywhere online. Not approximately. Character for character. A suite number present in one place and absent in another is a real inconsistency.
• A connected schema graph. Organization, LocalBusiness, Service, Person, and Article nodes with stable @id values that reference each other, rather than isolated blocks that share no relationships.
• A sameAs property listing your verified profiles: LinkedIn, Instagram, the directories where you are listed. This is how you tell search and AI systems which accounts belong to you.
• A Google Business Profile that agrees with the site exactly. Perplexity in particular leans heavily on Google Business Profile data for local recommendations.
• Named humans with credentials. Author attribution, real bios, and a team page. Models weight identifiable expertise, and an organization with no named people behind it is harder to vouch for.

Pillar 4: Get cited by sources the models already trust

This is the one that decides outcomes, and it is the one almost nobody does, because it is slow and it does not happen on your own website.

When a model answers best marketing agency near Seattle, it is not primarily reading agency websites. Every agency website claims to be the best; that claim carries no information. The model is synthesizing from sources with third-party credibility: directories, listicles, review platforms, forum threads, local press.

So the work is:

• Claim and complete the directories that already rank for your head terms. Search your own category and city, see which directories appear on page one, and get properly listed on those specifically. This has the useful side effect of putting you inside the assets that already own the SERP you cannot win directly.
• Participate where your category is discussed. Reddit threads, local Facebook groups, industry forums, and genuinely so, because a transparently promotional comment gets removed and does more harm than nothing. Reddit in particular is heavily represented in the training data and in live retrieval.
• Earn inclusion in local roundups and press. A mention in a regional publication or a well-linked local list is worth more than a dozen thin directory profiles.
• Build a real review base. Review platforms are heavily cited in AI local recommendations, and the language in your reviews becomes part of how the model describes you.

There is no shortcut here. It is a months-long campaign, and it is the difference between being technically optimized and being actually recommended.

How to measure whether any of this worked

AI visibility is measurable, but you have to build the measurement, because it does not appear in a standard analytics report.

1. Build a query bank. Ten to fifteen questions a real prospect would ask about your category and market. Not keywords, but full questions, phrased the way someone talks to a chatbot.
2. Run them on a schedule. Manually at first, or through the Perplexity and OpenAI APIs once you want a trend rather than an anecdote. Log whether you are named, where in the answer, and which competitors appear alongside you.
3. Segment AI referral traffic in GA4. Sessions from chatgpt.com, perplexity.ai, claude.ai, and gemini.google.com are identifiable and worth isolating from direct traffic.
4. Watch branded search volume. If AI recommendations are working, more people will search your business name directly. That shows up in Search Console.

The honest caveats

Three things worth stating plainly, because a lot of writing on this subject presents more certainty than exists.

Model behavior is unstable. Grounding strategies change without notice. A tactic that produces citations this quarter may not next quarter. Re-validate quarterly rather than assuming durability.

The engines differ substantially. Perplexity cites transparently and leans on directories and Google Business Profile data. ChatGPT cites inconsistently depending on whether browsing is triggered. AI Overviews behave differently again, and lean more on conventional search signals. Optimizing for all of them at once means doing the fundamentals well rather than chasing engine-specific tricks.

Nobody has confirmed mechanics. There is no documentation from any provider describing how citation candidates are selected. Everything in this article is inference from observed behavior, and it should be held that way.

Where to start this week

If you do only three things: check that your robots.txt is not blocking AI crawlers, add a direct-answer paragraph and an FAQ block to your three most important pages, and get properly listed on the two or three directories that already rank for your category and city.

That is a few hours of work and it moves you ahead of most of your competitors, because most of them have not started. If you want the full version, that is what our AI search optimization service covers, and it is the same method we run on our own site.