What Decides Whether AI Recommends a Local Business

WSI July 2026

Summary

Google local now runs as two parallel selection systems. AI Mode and the traditional Local Pack return different businesses for the same query, overlapping about 71%, and the criteria that get a business into one are not the criteria that get it into the other. Consumer adoption of AI for local discovery went from 6% to 45% in twelve months, and nearly all of those users verify the AI recommendation against a review profile before acting on it. The practical consequence is that rankings can hold steady while calls decline, and the standard local reporting most businesses receive will not show it.

Two things follow. The opportunity is unusually large, because adoption is climbing fast while most competitors are still measuring only one of the two systems. And the answer to what gets a business recommended is not the same answer in every category. The signal that lifts a plumbing company reverses for a law firm. Any playbook applied uniformly across verticals will work in some of them and quietly fail in others.

One reframing runs through the piece. Review flow now does acquisition work at two stages of the same purchase, which makes its usual home under reputation management a poor fit for what it's actually doing.

---

Key Highlights

• 28.5% of businesses appearing in the Local Pack were absent from AI Mode for the same query. Comparing top three results only, overlap fell to 48%. A further 999 businesses surfaced in AI Mode and never appeared in the Local Pack at all.

• AI local packs surfaced roughly 32% as many unique businesses as traditional three-packs, and showed fewer businesses in 88% of 322 markets tracked.

• Consumers using AI to find local businesses rose from 6% to 45% in a year, making AI the third largest local discovery channel. Google review readership fell from 83% to 71% over the same period.

• 97% of AI users cross-check AI recommendations against real reviews. AI decides the shortlist. The review profile decides the call.

• What separates the two systems inverts by industry. Plumbers appearing only in AI Mode carried 337% more reviews at the median. For lawyers and medical specialists, the pattern reversed and AI Mode more often returned educational content instead of business listings.

• Query phrasing alone can determine whether any business appears. A bare category term returned zero AI Mode listings across all ten cities tested, and adding "near me" returned listings in 38 of 38 cases.

• Reviews earned this month correlate more strongly with placement than lifetime review totals. An 18 day pause in review flow was enough to drop rankings sharply in one documented case.

• Click-to-call from Google Business Profiles has declined steadily across two years of tracked data while website clicks held, and paid local inventory expanded substantially over the same window.

• Advice to optimize the GBP Q&A section now points at a feature Google began removing in December 2025.

• The window is open. AI local packs currently run on roughly 7% of tracked keywords, mobile and US only, which is a thin competitive field against steeply climbing demand.

---

A year ago, 6% of consumers used an AI tool to find a local business. Today it's 45%.

That comes from BrightLocal's Local Consumer Review Survey, run across a representative panel of 1,002 US adults. It puts AI third among local discovery channels, behind only Google and Facebook, ahead of Yelp and Tripadvisor. Among 30 to 44 year olds the figure reaches 64%, which for most home services and professional services firms is the buying demographic. Over the same twelve months, the share of consumers reading Google reviews fell from 83% to 71%.

What makes that more than a channel-mix story is where those customers land. AI Mode and the traditional Local Pack are separate systems that return different businesses for the same query, and the monthly local report almost every business receives watches only one of them.

The engagement advice circulating right now isn't wrong. Post regularly, keep hours accurate, ask for reviews. Fine. But it's being offered as the answer to a question nobody has finished asking, which is whether the thing you're measuring is still the thing that decides whether customers find you.

The Number of Businesses Getting Surfaced Is Contracting

Sterling Sky worked with Places Scout to compare AI-powered local packs against traditional three-packs across their ranking reports. Traditional packs featured 18,330 unique businesses. AI local packs featured 5,943. That's roughly 32% as many businesses appearing at all.

Across 322 markets, 88% showed fewer unique businesses in the AI local pack than in the traditional one.

The mechanics behind that are straightforward. AI local packs show one or two businesses instead of three. They don't carry call buttons. And they surface a different set of businesses than the three-pack sitting below them.

Worth keeping the scale honest: Sterling Sky is seeing these packs on about 7% of the keywords they track, mobile only, United States only. That's a direction of travel with real numbers attached, some distance short of an emergency and some distance past a rumour.

The parallel trend is more immediately expensive. Jepto compiled click data across 179 Google Business Profiles at 34 US law firms over two years. Clicks to call declined steadily. Website clicks didn't. That split points at mobile, where the call button lives and where Google has been replacing it with images in several industries.

Meanwhile the paid surface expanded. Local pack ads appeared on 1% of Sterling Sky's mobile reports at the start of 2025 and about 22% by December. Local Services Ads went from roughly 11% of tracked queries to 31% over the same period.

That's the version of this that shows up in a business owner's revenue several months before it shows up in any report.

The Two Systems Select on Different Criteria

Local SEO Data ran 1,120 query executions between May 20 and 22, 2026, across seven verticals and 13 US cities, pulling both AI Mode and Local Pack results for each. Overlap between the two came in at 71.5%, with a 95% confidence interval of 69.2% to 73.8% across 10,000 bootstrap resamples. Put the other way, 28.5% of Local Pack businesses were missing from AI Mode for the same search. Compare only the top three from each and the overlap drops to 48%. AI Mode also surfaced 999 businesses that never appeared in the Local Pack at all.

What separates the two groups isn't consistent, and that's where it gets awkward.

In consumer trades, review volume tracks with AI Mode inclusion. Businesses appearing only in AI Mode had 49% more reviews at the median than businesses appearing only in the Local Pack. Among plumbers the gap reached 337%, a median of 546 reviews versus 125.

For lawyers and medical specialists, it inverts. Businesses in AI Mode had fewer reviews, and AI Mode more often declined to list businesses at all, returning educational content instead.

So the same optimization produces opposite results depending on the category. A plumbing company and a family law practice reading identical advice will get very different outcomes from following it, and neither will be able to tell why from their dashboard.

The researchers also found results shifting day to day. Broad patterns held across the collection window. Which specific businesses appeared for a given query did not.

Query Phrasing Determines Whether Any Business Appears

This finding deserves its own weight. In the same study, the bare query "divorce lawyer" returned zero business listings in AI Mode across all 10 cities tested. Adding "near me" brought listings back in 38 out of 38 cases.

Not fewer listings. None, then all.

That means visibility in AI Mode is partly a function of how the customer happens to phrase the question, independent of anything the business has done. A firm can be fully optimized, well reviewed, and completely absent because the searcher typed four words instead of six.

No business has a lever for that. What it does give you is a reason to stop treating a single tracked query as evidence of anything.

A firm can be fully optimized, well reviewed, and completely absent because the searcher typed four words instead of six.

Recency Carries More Weight Than Accumulated Totals

Sterling Sky analyzed 8,186 businesses across 200 cities, spanning large metros down to cities under 50,000 people, looking at what correlated with three-pack placement for "near me" queries in five service categories.

Reviews earned this month correlated more strongly than lifetime review count. They watched a dental client generating 60 or more reviews a month stop for 18 days, and the rankings dropped sharply. Competitors bringing in 15 to 45 a month held their positions through the same window.

Reviews containing written text correlated more strongly than star-only ratings, which fits with how a language model would use them. Review text describes what a business actually does. A number doesn't.

Two findings in that study cut against common advice. Hiding your address as a service area business, which Google's own guidance recommends, showed a negative correlation with ranking. Sterling Sky tested it directly on a home service client, watched rankings fall, restored the address a month later, and watched them recover. They repeated it on a second location with the same result.

And photos came back inconclusive. They help in visually driven categories like restaurants and salons. In categories like garage door repair, the same effort showed no measurable movement.

The prescription to upload photos twice a month, applied uniformly, is a real cost with a category-dependent return.

Parts of the Standard Checklist Point at Features That No Longer Exist

Google deprecated the Business Profile Q&A API on November 3, 2025, and began phasing out the public Q&A section from December 2025. New profiles are created without it. Existing content is frozen where it still displays.

The replacement is an AI-generated answer layer, surfaced through Ask Maps and powered by Gemini, that composes responses from the profile, the reviews, the photos, and the website.

Tactics outlive the products they were built for, and they keep circulating because they still sound reasonable.

Advice to seed your Q&A with five customer questions was sound in 2024. Today it points at a control panel that isn't there. And the answers customers receive are now assembled from source material the business influences indirectly rather than writes directly.

This is the recurring shape of the problem. Tactics outlive the products they were built for, and they keep circulating because they still sound reasonable. The checklist stays stable while the surface underneath it moves.

What the Ranking Factor Research Actually Establishes

The Whitespark 2026 Local Search Ranking Factors report is the most cited document in local search, and it's worth being precise about what it is. Darren Shaw surveyed 47 local SEO practitioners and published their aggregated weightings in November 2025. It reflects expert judgment about influence. It isn't a measurement of causation.

Read that way, it's genuinely useful. Google Business Profile signals carry the heaviest weight of any category at roughly a third of the total. Reviews land around a fifth. Citations, which absorbed a decade of agency effort, have declined to single digits. AI search visibility appeared as a formal category for the first time.

Primary category ranks as the single most influential individual factor. Being open at the moment of search ranks fifth, a finding Joy Hawkins surfaced and BrightLocal tested across 50 businesses in 10 categories.

A survey of expert opinion and a controlled study aren't the same evidence, and treating them as interchangeable is how the industry ends up confidently wrong. The distinction matters more now, not less, because the surface is changing faster than anyone's ability to measure it cleanly.

AI Recommendations Get Verified Before They Get Acted On

The same BrightLocal research found something that reframes what the two systems are each doing.

Among consumers who use AI for local recommendations, 88% check the information they're given, either verifying that a cited review is legitimate or looking at the source. And 97% said they sometimes double-check AI recommendations against real reviews, with 42% always going to a native review platform before deciding.

So the customer journey has picked up a stage rather than swapping one surface for another. AI Mode or ChatGPT produces the shortlist, and then the review profile gets pulled up to confirm it. Trust in the recommendation is high, at 63% among AI users, and the verification behavior happens anyway.

That changes what each surface is responsible for. Getting named in the AI answer is what puts you on the list. Whether you stay on it depends on what the customer finds thirty seconds later when they go and look you up. A business can win the first stage and lose the second, and the only visible symptom is a phone that rings less than the rankings suggest it should.

It also explains why review recency keeps outperforming review totals across studies that share no methodology. A recency-weighted profile happens to satisfy two very different readers at once, the model composing the answer and the person auditing it.

My read on that is a budgeting argument more than a marketing one.

Reviews are funded as cleanup and performing as demand capture.

Reviews have sat under reputation management for as long as most businesses have had a Google listing. That's a defensive category. It carries a small line item, it usually reports to customer service or an office manager, and it gets attention when something goes wrong.

The sequence in this data puts review flow somewhere else entirely. If the AI answer assembles the shortlist partly from review patterns, and the customer then confirms that shortlist against the review profile, reviews are doing acquisition work at two separate stages of the same purchase. Under that description they're funded as cleanup and performing as demand capture.

The practical version is unglamorous. Whoever owns review generation should sit closer to the people accountable for pipeline than to the people handling complaints, and the budget line should reflect a channel rather than a maintenance cost. Most mid-market businesses I see have that reversed, and the reason is historical rather than deliberate. Reviews were a reputation problem for fifteen years. The category label outlived the job.

The Same Signal Moves in Opposite Directions by Vertical

Read the findings above in sequence and a pattern shows up that none of them states individually. Review volume separates the two systems in consumer trades and reverses in professional services. Photos correlate in restaurants and salons and do nothing measurable in garage door repair. The bare query that returned no AI Mode listings at all was a legal service, not a trade.

Two studies, different methods, different samples, and the direction of the effect keeps flipping on category.

That has an uncomfortable implication for how local marketing is usually sold. A checklist that recommends the same eight actions to every business is describing an average that almost no individual business sits at. The plumbing company and the family law practice can follow it with equal discipline and get opposite results, and neither will find the reason in their reporting.

The more useful framing is to ask what the AI answer looks like in that category before deciding what to fix. Where the answer names businesses, the work is roughly what the local playbook always said, sharpened by recency: sustained review flow, real photographs, an address Google can see. Where the answer explains rather than lists, the profile isn't the contested asset. The contest moves to whose material gets cited inside the explanation, and no amount of review volume resolves it.

A checklist that recommends the same eight actions to every business is describing an average that almost no individual business sits at.

Ten minutes of running a client's actual commercial queries through AI Mode answers that question. Most local marketing plans are written without it.

This is also where the opportunity sits, and it's a large one. That 7% keyword footprint means the competitive field is thin at exactly the moment consumer adoption is climbing steeply. The 999 businesses appearing in AI Mode and nowhere else are a reasonable description of what an unclaimed position looks like.

Windows like this one close, and while they're open they don't reward generic effort. They reward knowing which of these signals the business in front of you actually runs on.

The Measurement Gap Is the Actual Exposure

Every finding above shares a structural feature. The signal that would tell you something's wrong doesn't reach the report you're reading.

Calls fall while position holds. A Local Pack ranking looks healthy while an entirely separate set of competitors occupies AI Mode results nobody on the account has ever pulled.

Review velocity slips for two weeks and the consequence lands a month later. A query phrasing you never tracked decides whether you exist in one of the two systems.

All of that is instrumentation rather than content or effort, and adding more activity on top of instrumentation you can't trust produces motion without information.

The practical starting point is narrow. Pull AI Mode results for your top five to ten commercial queries alongside your existing Local Pack tracking, and run them repeatedly rather than once, because the results move. Test each keyword bare and with "near me," since that pairing is the one the study measured and the gap it found was total. City name and "best" are worth adding on the same principle. Separate your Google Business Profile website clicks from general organic traffic with UTM parameters so a decline in one doesn't hide inside the other. And watch reviews per month as a rate, not reviews total as a milestone.

That's a measurement build, not a campaign. It costs a fraction of what most businesses spend on local marketing, and until it exists, the rest of the spend is being evaluated against a picture that's missing about a third of the frame.

Adding more activity on top of instrumentation you can't trust produces motion without information.

The timing argument is simple arithmetic. A channel went from 6% to 45% of local discovery in twelve months. The reporting most local businesses receive still describes the channel it displaced.

The businesses that adapt well here won't be the ones posting most often. They'll be the ones who can tell, within a week, that something changed, and who know which signals their category actually runs on.

---

Frequently Asked Questions

Is AI Mode replacing the Google Local Pack?

Not currently. They operate as parallel systems returning different results for the same query. Research from Local SEO Data found about 71.5% overlap between the businesses appearing in each. Sterling Sky reports seeing AI-powered local packs on roughly 7% of tracked keywords, mobile only, in the US. The two coexist, and tracking one tells you little about the other.

Why does my business appear in Google Maps but not in AI Mode?

Because they select differently. In consumer service categories, businesses appearing only in AI Mode carried substantially more reviews at the median than Local Pack businesses, 49% more overall and 337% more among plumbers. In legal and medical categories the pattern reverses and AI Mode frequently returns educational content instead of business listings. Query phrasing also matters. Some bare keywords return no listings at all until a location modifier is added.

What happened to the Q&A section on my Google Business Profile?

Google deprecated the Q&A API on November 3, 2025 and began removing the public Q&A section from December 2025. New profiles don't include it. Customer questions are now handled by an AI-generated answer layer, delivered through Ask Maps, that draws on your profile details, reviews, photos, and website content. Moving your old Q&A content onto your own site is the practical response, since that's a source the AI layer reads.

Do Google Business Profile posts improve rankings?

The evidence for posts is considerably weaker than the evidence for reviews. Behavioral and engagement signals appear in the Whitespark 2026 rankings survey, but that report reflects surveyed practitioner opinion rather than controlled measurement. Where studies have isolated variables, review recency and review text show clear correlation with placement. Posting is inexpensive and unlikely to hurt. Treating it as a primary ranking lever isn't supported by the same quality of evidence.

Do business hours affect local rankings?

Yes. Being open at the time of search ranks as the fifth most influential individual factor in the Whitespark 2026 survey. Joy Hawkins of Sterling Sky first surfaced the pattern, and BrightLocal tested it across 50 businesses in 10 categories, finding rankings tended to fall when a business was listed as closed. Holiday hours and answering service coverage are worth auditing on a schedule rather than reactively.

How many Google reviews should my business get each month?

Rate matters more than total. In Sterling Sky's analysis of 8,186 businesses, reviews earned in the current month correlated more strongly with placement than lifetime count. Their case data showed competitors sustaining 15 to 45 reviews per month holding steady positions, while a practice that paused for 18 days saw a sharp drop. Reviews with written text correlated more strongly than star-only ratings.

How many people actually use AI to find local businesses?

BrightLocal's 2026 Local Consumer Review Survey, based on a representative panel of 1,002 US adults, found 45% had used an AI tool for a local business recommendation in the past twelve months, up from 6% the year before. ChatGPT led at 31%, with Google's AI Mode at 23%. Adoption peaks among 30 to 44 year olds at 64% and falls to 24% among those over 60. Over the same period, consumers reading Google reviews dropped from 83% to 71%.

Do people trust what AI recommends for local businesses?

Trust is high and verification is near universal. Among AI users, 63% trust AI recommendations and 88% check the information they receive. 97% said they sometimes cross-check AI recommendations against real reviews, and 42% always visit a native review platform first. Practically, AI decides who makes the shortlist and the review profile decides who gets the call.

Should a service area business hide its address on Google?

Google's guidance says yes. The measured outcomes point the other way. Sterling Sky's study found a negative correlation between hidden addresses and "near me" ranking, and their direct test showed rankings falling when an address was hidden and recovering when it was restored, replicated across a second location. This one carries policy risk, so it's a decision to make deliberately rather than by default.

---

Sources

Sterling Sky. "The State of Local SEO in 2026." June 2026.

Sterling Sky. "We Analyzed 8,186 Businesses in 200 Cities. Here's What Actually Gets You Ranking for 'Near Me' in 2025." November 2025.

Local SEO Data. "AI Mode vs Local Pack: Where Do Local Businesses Actually Appear?" June 2026.

Whitespark. "Local Search Ranking Factors 2026." November 2025.

BrightLocal. "Nearly Half of Consumers are Asking AI for Business Recommendations." March 2026.

BrightLocal. "Local Consumer Review Survey 2026." February 2026.

BrightLocal. "Study: Business Opening Hours and Local Rankings."

Google Business Profile Q&A API deprecation. November 2025.