How Google’s AI Decides Which Local Business to Recommend

Google set out, in guidance aimed at small businesses, what its AI considers when deciding which local businesses to surface. It is a short list and nothing on it is surprising, which is itself the useful finding.

The five

• Upload images and video regularly. Media gives the model context that text fields cannot. A profile with current job photos is legible in a way a text-only listing is not.
• Fully complete the profile. Services, pricing, specialisms, attributes. The framing Google uses is blunt: if the information is not on your profile, the AI cannot recommend you for it. You are not being outranked on those queries, you are absent from them.
• Accumulate reviews, and respond. Volume and recency both count, and responses are part of the signal rather than a courtesy.
• Make sure Google can actually read your site. Content that is hidden behind interactions, rendered late, or locked inside images cannot be used. If a crawler cannot see it, no AI feature can cite it.
• Keep business information consistent everywhere. Website, profile, directories. Conflicting data across sources reduces confidence, and low confidence means you get left out rather than ranked lower.

The line worth reading twice

Inside the guidance on crawlability sits a qualifier that undercuts a common tactic: use structured data only when it matches what people can actually see on the page.

Marking up content that is not visible to a human visitor (invisible FAQ blocks, review schema for reviews not shown, service lists that exist only in JSON-LD) is not a shortcut into AI results. It is a guideline violation, and it has been for a long time. The AI era has not created an exemption.

Related: Google's AI optimization guide states that Google Search ignores llms.txt files and other AI-specific markup entirely. Between those two statements, most of the "technical AI SEO" product category has very little left to stand on.

What this means if you are a contractor

Two of the five are about completeness and two are about consistency. Only one (regular media) asks for ongoing creative effort, and it is satisfied by photographing jobs you are already doing.

The pattern across everything Google has published on AI and local this year is the same: the profile is the asset, completeness is the lever, and there is no separate AI channel to buy your way into.

That is worth knowing mostly as a defence. When someone pitches you an AI visibility retainer, you can now ask which of these five it improves, and what it does that filling in your own profile would not.

What the data actually shows

To understand how well contractors are currently meeting Google's own criteria, we pulled 1,440 Google Business Profiles (the top-10 results across 29 US cities and five trades) in May and July 2026. What we found is that most businesses appearing in top-10 results are succeeding despite incomplete profiles, not because of complete ones.

The services section is the starkest example. Across all 1,440 profiles collected in that period, 1,426 had no services section populated at all. That is not a rounding error. In Salt Lake City alone, 49 of the 50 top-10 profiles across the five trades had nothing in the services section. Google's own guidance says the AI cannot recommend you for something not listed on your profile. Most top-ranked contractors are sitting in a blind spot they do not know exists.

The picture is consistent by trade. Among roofing profiles in our sample, 288 of 290 had no services section, with a median review count of 172. HVAC profiles showed 287 of 290 with no services populated and a median of 930 reviews. Plumbing came in at 287 of 290 with a median of 728 reviews. Electrical was 285 of 290 with a median of 134 reviews. Landscaping was 279 of 280 with a median of just 80 reviews.

The review counts matter here because they suggest the mechanism these businesses have been relying on: volume of reviews has been doing enough work to get them into the top 10 even without completeness signals. That may hold for standard search ranking. It is less likely to hold as AI-driven recommendations become more selective, because those features depend on understanding what a business actually does, not just how many people have reviewed it.

Other signals from the same dataset reinforce the picture. Sixty-five percent of the profiles in our sample had no booking link. Every single profile, all 1,440 of them, had gone more than 90 days without a post. Ninety-three percent had no Q&A entry. Four percent had no website linked at all. These are not fringe cases. They are the norm among businesses currently winning top-10 placement.

The implication is not that these businesses are about to fall. It is that the gap between current profile state and Google's stated requirements for AI recommendation is wide open, and the businesses that close it first will have a straightforward structural advantage over competitors who do not.

How to actually act on this: a step-by-step approach

The five factors Google identified are not equally time-consuming. Some can be addressed in an afternoon; others require a habit rather than a single session. Here is a sequenced approach based on effort and likely impact.

1. Audit your services section first. Log into your Google Business Profile and navigate to the services tab. If it is blank (and based on what we found in our May and July 2026 data collection, there is a high probability it is), this is your highest-priority task. Add each service you offer as a discrete item. Write a plain description for each one: what it includes, what conditions you work in, what types of property or equipment you handle. Do not stuff keywords. Describe the service the way you would to a customer calling for the first time.
2. Check every factual field for consistency. Your business name, address, phone number, and hours should be identical across your GBP, your website, and any directory listings you have claimed. Pull them up side by side. Inconsistencies that look minor (a suite number abbreviated differently, a phone number with and without an area code) are enough to create the kind of conflicting signal Google's guidance flags as a confidence reducer.
3. Make your website content visible to a crawler. This one requires either a developer or a tool that can render your site the way Google does. The practical check: if you disable JavaScript and reload your service pages, can you still read the content? If the page is blank or broken, Google's crawler may be seeing the same thing. Content in PDFs, image overlays, or loaded by user interaction after the initial page render is not reliably indexed.
4. Build a photo habit, not a photo project. A one-time upload of 40 photos from three years ago is less useful than four photos added this week and four more added next week. Take pictures of jobs as you finish them. Before-and-after pairs are useful because they show scope. Equipment photos are useful because they signal specialization. You do not need a photographer. A phone with decent light is sufficient for this purpose.
5. Set a review response schedule. Responding to reviews is listed as part of the signal, not a bonus. The practical approach is to block 15 minutes once a week to respond to anything that came in. Responses do not need to be long. They need to be present, specific enough to reference what the reviewer mentioned, and consistent in tone.
6. Answer the booking questions somewhere Google still reads. The Q&A field was removed in December 2025 and Ask Maps replaced it. Ninety-three percent of the profiles in our sample had none. Those answers now belong in your service descriptions and on your website. Pick a question you actually get on the phone, something like service area, whether you offer free estimates, or which brands you work with. This is a thin signal but it is one that almost no competitor is using.

Who this does not apply to

This guidance is specific to businesses that show up in local map results: trades, services, and other businesses where location is a primary search filter. If you run a fully remote business with no physical service area, Google Business Profile optimization is not your lever. Similarly, if you operate in a market where your category has very low search volume, the competitive dynamics are different enough that the priority order above may not hold.

This also does not address paid local advertising, Local Services Ads, or social platforms. Those are separate channels with separate ranking logic. The factors described here apply specifically to organic Google Business Profile visibility and the AI-driven recommendation surfaces that draw from it.

Frequently asked questions

Does filling in the services section actually change where I rank in standard local search?

The services section has a direct effect on which queries your profile can surface for, because Google's systems use it to match your business to searches. Whether it changes your rank within a query you were already appearing for depends on other factors. The more immediate effect is inclusion versus exclusion: a roofing company with no services listed may not appear for "flat roof repair" even if it is the closest and best-reviewed option in the area.

How many photos is enough?

Google has not published a specific threshold, and any number you see cited without a named source should be treated as speculation. What the guidance emphasizes is recency and regularity rather than a one-time count. A profile that adds photos consistently signals an active business. A profile with 200 photos uploaded once in 2021 and nothing since does not carry the same signal.

If I already rank in the top three, do I still need to do this?

Your current position reflects how the ranking system has weighted signals to date. AI-driven recommendation features (AI Overviews, the AI Mode local pack, voice-based results) weight completeness signals differently than the standard map ranking algorithm does. Say you run a plumbing company and you rank second in the standard local pack without a services section. That position may hold for standard results while you remain absent from AI-generated recommendations that pull specifically from structured profile data. Treating them as the same channel is increasingly not accurate.

Is there any point posting to GBP if every profile in your sample had gone 90 days without one?

The fact that 100 percent of the 1,440 profiles we collected in May and July 2026 had gone more than 90 days without a post tells you what the current competitive floor looks like, not what the ceiling is. Google's guidance includes posts as a freshness signal. In a competitive market where no one is posting, a business that posts consistently is doing something none of its visible competitors are doing. Whether that translates to a meaningful ranking shift depends on many other factors, but it costs very little to find out.