AEO for service businesses: how AI picks who to recommend

Type “best remodeler in Denver” into ChatGPT and you get a short, confident list of names. Type “plumber near me” into Google and — more often than not — you still get the familiar map with three pins, no AI answer in sight. Both things are true at once, and the gap between them is the whole story of AEO for local service businesses.

The headline most local owners have heard is “AI is eating search.” For service businesses, the reality is more specific and more useful: AI mostly stays out of the moment when someone is ready to hire — and dominates the moments right before it. Understanding which queries AI answers, and what it reads when it does, is the difference between chasing a threat that isn't there and winning the one that is.

01The local AEO paradox

Start with where AI actually shows up. Ahrefs analyzed 146,122,391 desktop search results and found that AI Overviews trigger on just 7.9% of local queries versus 22.8% of non-local ones.¹ seoClarity, using a stricter definition of “local” across 500M+ keywords, put the figure closer to 0.01% — while AI Overviews hit roughly 30% of US desktop keywords overall by September 2025, up from 10% that March.² However you slice it, the local pack still owns the transactional search.

But averages hide the mechanism. Whitespark ran 540 queries across Denver, Houston, and Phoenix in six service industries and split them by intent. The result is the most important chart a service business can see this year:³

That's the paradox: the query that converts — “plumber near me” — is the one AI mostly ignores. The queries that shape the decision — “how do I choose a remodeler,” “best design-build firms in Denver,” “Angi vs Houzz for contractors” — are the ones AI answers 92–97% of the time. By the time your buyer opens the map, AI has already handed them a set of names to look for. If you're not in that set, you're competing to be chosen from a list you were left off of. And this matters more every quarter: 40% of consumers now say they use generative AI when searching online.⁴

02When AI does answer local, directories decide

Here is where local AEO stops looking like normal SEO. When AI assembles a local answer, it overwhelmingly reads directories and review platforms, not the businesses themselves. In Whitespark's study, roughly 60% of the citations behind a local service answer pointed to third parties — Yelp, Thumbtack, HomeGuide, Reddit, Indeed — and only about 40% to individual local businesses.³ Foundation put hard numbers on it, analyzing more than 28 million AI responses:⁵

The lesson isn't “go pay Yelp.” It's that AI answers your category by reading the places that aggregate your category. Your own website is a supporting citation; the directory profile is the primary one. SE Ranking, modeling 129,000 domains, found that having profiles on Trustpilot, G2, Capterra, Sitejabber, and Yelp gave a domain 3× higher odds of being cited by ChatGPT.⁶ For a service business, the equivalent list is your Google Business Profile plus the two or three directories that own your trade — and being genuinely present on them, not just claimed.

03Reviews win the recommendation, not code

If directories are where AI reads, reviews are what it reads for. Insites analyzed 10,000 local businesses against ChatGPT and Perplexity recommendations and found a stark separation: the businesses AI recommended averaged 133 Google reviews; the ones it ignored averaged 11. Local Pack presence made a business 3.5× more likely to be recommended. And the factor with the weakest correlation of everything they measured was technical SEO — Core Web Vitals, structured data, the stuff agencies sell hardest.⁷

Seer Interactive saw the same shape at scale. Across 804,491 AI responses, brands with no review profile had a median AI-citation rate near 1%; brands with even 1–13 reviews jumped to 53.5%.⁸ The first thirteen reviews matter more than the next hundred. This is the most encouraging finding in local AEO: the bar to become visible is low, boring, and entirely within your control — you just have to actually clear it.

04The framework you already have to win

Google has published its local ranking logic for years, and it maps almost perfectly onto what AI rewards. The three signals are relevance, distance, and prominence— and Google states plainly that “more reviews and positive ratings can help your business's local ranking.”⁹ Whitespark's 2026 survey of local search professionals weights those signals into the factor groups below (reviews rose from 16% of the weighting in 2023 to 20% today):¹⁰

Read the stack top to bottom and the strategy writes itself. Distance you can't change. Relevance you set correctly once — the right primary category, honest service listings, city named on the page. Prominence is the ongoing work, and it's the layer both Google's pack and the AI models weight most: reviews and citations. Spend your effort where it compounds.

05Reviews: recency is the lever, not just the count

“Get more reviews” is half the advice. The other half is keep them coming. In BrightLocal's 2026 Local Consumer Review Survey of 1,002 US adults, 97% read online reviewsfor local businesses, and 47% won't use a business with fewer than 20 of them. 31% will only consider a business rated 4.5 stars or higher.¹² But the number that should change your process is about time:

• 32% only look at reviews from the last two weeks — up sharply from 20% a year earlier — and 74% want reviews from the last three months.¹² A wall of five-star reviews from 2023 reads as stale to buyers and models alike.
• Google scores reviews on a rolling average of all published 1–5 ratings, and a new review can take up to two weeks to affect the score — so recency is something you build continuously, not in a burst.⁹
• 85% say a business's response to reviews makes them more likely to use it, and 81% expect a reply within a week.¹² Responding is free reputation work that both audiences see.

06What to do this quarter

None of this requires a rebuild. In priority order:

• Set your Google Business Profile category correctly — the single heaviest local factor, and free. Pick the most specific primary category that fits, and fill every relevant service.^10,11
• Get on the two or three directories that own your trade — for most services that starts with Yelp, plus the vertical directory AI reads in your category (Angi, Houzz, Thumbtack). AI cites the aggregators first.^5,6
• Turn reviews into a monthly habit, not a campaign. Clear the ~20-review floor, hold 4.5+ stars, and keep fresh ones landing inside the last two weeks. Reply to them.^7,8,12
• Stop over-investing in technical SEO for AI. Crawlable and fast is table stakes; it showed the weakest correlation with being recommended. The returns are off-site.⁷
• Measure the questions your buyers actually ask. Run the “best [service] in [city]” and “how do I choose” queries across the models and see whether you're named.

One caution on that last point: local AI answers are volatile. SparkToro ran 2,961 prompt trials and found less than a 1-in-100 chance that the same question, asked 100 times, returns the same brand list in any two responses.¹³ You can't judge your visibility from a single lucky query — you have to track the pattern over time. But the underlying signals AI reads to build those lists are stable, unglamorous, and yours to build: the right category, the directories that own your trade, and a review profile that never stops growing. The businesses that start now are the ones AI defaults to later.