How Do You Get Cited by AI?

What does "cited by AI" actually mean?

It means that when somebody asks an AI assistant a buying question — who's the best plumber in Santa Rosa? — your business is one of the names that comes back.

That is not the same as ranking on Google, and it is not traffic. It is being named in the answer. In a world where the answer increasingly replaces the click, being named is the thing worth having.

Why does this matter more than it used to?

Because a growing share of searches never produce a visit to anybody's website. Research from SparkToro using Similarweb clickstream data found 68.01% of US Google searches ended without a click in the first four months of 2026, up from 60.45% in 2024. Google's AI Overviews appear on roughly a fifth of searches, and where they appear, click-through drops sharply.

We can show you the same thing from our own site. Last month baadigi.com recorded 68,764 impressions and 58 clicks — a click rate of 0.08%. One page had 57,322 impressions and one click. We were seen constantly and visited almost never.

What actually gets a business cited?

In August 2026 we checked 91 businesses to see whether AI assistants named them for their trade and city. 70 were named. 21 were not. Comparing the two groups is where the useful patterns are.

Third-party corroboration beats self-description

The businesses that got named tended to appear in places they do not control — Google Business Profile, reviews, local directories, industry listings, roundup articles. The ones that did not tended to exist mainly on their own domain. AI assistants are pattern-matching across sources; one source is a claim, several is a fact.

Reviews are doing heavy lifting

Review volume and recency came up repeatedly among the named businesses. This is not a mystery: reviews are third-party text, they mention the service and the location naturally, and they are constantly refreshed. They are close to the ideal source material.

Category words matter more than brand words

This one surprised us on our own site. Pages can answer a question perfectly and still not get cited, because they answer it in the wrong vocabulary. A page about "roofing lead costs" that never uses the phrase buyers actually type will lose to a weaker page that does. Write in category language, not internal language.

A working website is the floor, not the lever

Here is the finding that reframed this for us. Of the 21 businesses that were not named, 12 had technically clean websites — they passed every agent-readiness check we ran. Their sites were not the problem. Rebuilding them would not have fixed anything.

Conversely, 30 businesses were named by AI despite having real defects on their sites.

So a broken site can undermine you once someone arrives, but fixing it does not, on its own, get you named. These are two different jobs and they are frequently sold as one.

What should you do first?

1. Find out where you actually stand. Ask an AI assistant the question your customers would ask. Not your business name — the category question. See who comes back.
2. Fix your Google Business Profile properly. Correct categories, real service list, accurate hours, genuine photos. It is the most consistently referenced source about a local business.
3. Build review velocity, not a review total. A steady trickle beats a burst two years ago.
4. Get listed where your industry gets listed. Trade bodies, licensing directories, chambers, reputable local directories.
5. Earn a mention you did not write. Local press, supplier case studies, community sponsorship, an industry roundup. One genuine third-party mention outperforms a page of self-description.
6. Then check your site can be used — see can AI agents use your website and what an llms.txt file does.

How do you know it's working?

Re-run the same questions on a schedule and write down what you see. Monthly is sensible.

One caution worth taking seriously: AI answers vary between runs. We saw meaningfully different results asking the same question minutes apart. Never treat a single check as a verdict — look for the pattern across repeated checks and over weeks.