How to Get Recommended by ChatGPT

There is no submit form. There is a set of signals, and most of them are off your website.

How The Recommendation Gets Made

When someone asks an AI assistant for a recommendation in a specific place, two different mechanisms can produce the answer.

The first is retrieval. The assistant runs a web search, reads the top results, and writes an answer from what it found with citations. Most local recommendations work this way, because the model knows its own training data is stale and locations change.

The second is the training data itself. If a business is well known enough to have appeared repeatedly in the text the model was trained on, it can be named without any lookup. For a national brand this matters. For a local business it almost never does.

That split tells you where to spend effort. Since your recommendation comes from retrieval, it comes from what is findable about you on the open web right now. Which makes this much less mysterious than it sounds.

Rank First, Get Cited Second

Because retrieval starts with a search, being absent from search results makes you absent from the answer. There is no back door.

So the first requirement is unglamorous. Rank for the local queries that matter, keep your Google Business Profile complete and current, and make sure the pages that describe your services exist and are indexed.

Anyone offering to get you into ChatGPT without touching your search presence is selling something that does not work. The whole pipeline runs through search.

Third Party Sources Carry Unusual Weight

Here is where it differs from ordinary SEO. When these systems answer a recommendation question, they lean heavily on sources that are not the business's own website. Directories, review platforms, local news, roundup articles, forum threads, trade association listings.

The reason is straightforward. Your own site claims you are the best plumber in town, and so does every other plumber's site, so that claim carries no information. A third party saying it does.

The practical consequence is that a good deal of answer engine work happens off your website. Get listed accurately in the directories that matter in your trade. Get on your local chamber and trade body pages. Be the business a supplier lists on their find an installer page. Where local publications run best of roundups, be in them.

This is the same reputation work that has always made local businesses findable, and it now pays twice.

Reviews Are Read As Text

A retrieval system reading your review pages reads what people wrote, and the star count is only the start.

This matters more than it sounds. If someone asks for a dentist who is good with nervous patients, the assistant is looking for evidence of that specific thing, and the evidence is in review text. A business with forty reviews that repeatedly mention patience with anxious patients will be named. A business with two hundred reviews that all say great service will not.

You cannot script what customers write and should not try. You can ask for reviews at the moment the specific thing went well, and you can reply to reviews in a way that names the service and the area, since your replies are text on the page too.

Volume, recency and spread across platforms all matter. Sixty reviews across Google and two trade platforms is a stronger signal than sixty on Google alone.

Make Your Own Pages Extractable

Your website still has a job, which is to confirm and add detail once the assistant has found you.

The things that survive extraction are specific facts. Areas covered, named as actual places. Services, named as the things people call them. Prices or ranges. Hours including exceptions. Qualifications and registration numbers. Response times.

The things that do not survive are the things most local sites are made of. Claims about quality, mission statements, and history. Nothing in "we pride ourselves on outstanding customer service" can be quoted usefully, because every competitor says it.

Check whether your site is readable at all. Some assistants fetch pages directly, and a site that renders its content through JavaScript may return an empty page to a fetcher. Load your page with JavaScript disabled and see what is there.

Check What They Say About You

This is the step almost nobody does, and it is the one that produces the surprises.

Write down the fifteen questions a customer might ask before hiring you, including the ones naming your town and the ones comparing options. Ask them in ChatGPT, Perplexity, Gemini and Google. Record three things: whether you appear, who appears instead, and whether what is said about you is correct.

The last one is where the value is. We regularly find businesses being described with hours they no longer keep, services they stopped offering, or an old address, all traced back to a stale directory listing that nobody had thought about in years. Fixing the source fixes the answer.

Do this monthly. It takes half an hour and it is the only honest measurement available, because none of these platforms provides anything like Search Console.

If you would rather not run it by hand, this is part of what askotter monitors on the sites it manages, alongside the search and analytics data. Self serve starts at $99/mo, managed is $600/mo.