Search became an answer.
Your business needs to be in it.

Ask ChatGPT who the best roofer in your city is. You will get a name, a short justification, and maybe two alternatives. You will not get ten blue links and an invitation to research. The answer is the result.

That shift has a measurable footprint. Similarweb found that zero-click searches on Google — searches that end without the user visiting any website — climbed from 56% to 69% in the year following the rollout of AI Overviews. Roughly seven in ten searches now resolve on the results page itself.

The traffic you lose is not the traffic that mattered

The instinct is to read that number as catastrophic. It is more complicated than that, and the more interesting finding cuts the other way.

Seer Interactive measured conversion rates by traffic source and found that visitors arriving from ChatGPT converted at 15.9%, compared with 1.76% for traditional organic search. That is roughly a nine-fold difference. It makes sense: the AI has already done the filtering. By the time someone clicks through, they are not comparison shopping — they have been handed a recommendation and they are acting on it.

So the picture is fewer visitors, far better ones. Which reframes the whole problem. The question is no longer "how do I rank third instead of eighth." It is: when the assistant names one business, is it mine?

What actually drives an AI citation

This is where GEO — generative engine optimization — stops being a buzzword and starts being an engineering problem. A few findings have held up well enough to build on.

Statistics and citations outperform adjectives

A team from Princeton, Georgia Tech and IIT Delhi published the first systematic GEO study at KDD 2024, testing which content changes increased the likelihood of being cited in a generated answer. Adding relevant statistics to content improved visibility in AI responses by roughly 41%. Adding quotations and citations produced similar gains. Keyword stuffing, the old reflex, did essentially nothing.

The mechanism is intuitive once you see it. A language model assembling an answer needs quotable, verifiable material. "We provide exceptional service" is unquotable. "Average response time of 43 minutes across 1,200 service calls in 2025" is a sentence a model can lift and attribute.

Mentions beat backlinks

Ahrefs analyzed the relationship between various signals and AI visibility and found brand mentions correlated at 0.664, while backlinks correlated at just 0.218. Being talked about matters more than being linked to.

This is the hardest pill for anyone who has spent a decade optimizing their own site. Most of your AI visibility is not built on your domain. It is built on directories, local news, industry roundups, community forums, supplier pages, association listings, and review platforms. Your website establishes the facts; the rest of the internet establishes that you exist.

Consistency is an accuracy problem

Language models are pattern matchers. If your business name, service list, address and hours differ across four directories, the model has four competing versions of you and low confidence in all of them. Low confidence means it recommends someone else. Entity consistency is unglamorous and it is table stakes.

Five moves that matter for a local business

1. Baseline it first. Write down the ten prompts a real customer would use — "best HVAC company in Katy TX," "who fixes slab leaks near me," "emergency plumber open Sunday Houston" — and run every one through ChatGPT, Gemini, Perplexity and Google AI Overviews. Record who gets named. That list is your competitive set, and it is often not the competitors you assumed.
2. Make your facts machine-readable. Services, service area, pricing structure, hours, credentials, and warranty terms stated plainly on your own site, marked up with structured data. Not buried in a hero image or a PDF.
3. Publish something quotable. Real numbers from your own operation. Response times, job counts, years in business, warranty terms, seasonal cost ranges for your market. This is the single highest-leverage content change available to a local business.
4. Build earned mentions deliberately. Chamber listings, trade association directories, supplier partner pages, local press, sponsorships, and genuinely useful answers in community forums where people ask for recommendations.
5. Feed the review corpus. Assistants read review text, not just the star average. Reviews that name specific services and specific neighborhoods teach the model what you do and where. Generic five-star reviews teach it almost nothing.

The honest caveat

GEO is young. The measurement tools are immature, the platforms change their retrieval behavior without notice, and anyone selling you a guaranteed position inside ChatGPT is selling something they cannot deliver.

What is not in doubt is the direction. Similarweb's 2026 data shows 35% of U.S. consumers already using AI for product discovery. The businesses building entity clarity, earned mentions and quotable content now are compounding an advantage that will be expensive to buy later — the same way the businesses that took Google seriously in 2004 spent the next decade being very hard to catch.