Open your Google Business Profile this week and click the Performance tab. Google added a row in 2026 that was not there last spring. It splits your impressions across three surfaces: the classic local panel, the Maps AI summary card, and the AI Overview that sits at the top of the organic results page. That third number decides whether your business shows up when a customer asks Gemini a question instead of scrolling a list.
The shift is not subtle. On searches where an AI Overview appears, clicks to the websites underneath drop by about 58 percent. The customer reads the answer and never scrolls. So the question stopped being “where do I rank” and became “is my business one of the few Google quotes inside the answer.” Your new AI-surface share row is the scoreboard for that second question, and most local profiles are losing it without knowing the row exists.
What the AI-surface row actually measures
The old Insights tab became the Performance tab back in 2023. The 2026 overhaul went further. It now breaks every impression into where it happened. Classic local panel is the familiar three-pack and the map. The Maps AI summary card is the answer box that appears when someone asks Maps “who near me sells X.” AI Overviews is the cited answer block in regular Google Search. Three surfaces, three counts, one new ratio that matters.
That ratio is your AI-surface share: the percentage of total views coming from the two AI surfaces rather than the classic panel. Industry tracking puts a working floor at 15 percent. If your share sits under 15 percent, one of two things is true. Either your profile is missing the grounding content these surfaces read, or you operate in a vertical where Google has not switched on summaries yet. The fix depends entirely on which one it is, so check before you spend a dollar.
A quick example shows why the row matters more than the total. A two-location dental practice in Boise watched total profile views fall 22 percent over the spring and assumed something broke. The Performance tab told a different story. Classic-panel views dropped hard, but the Maps AI summary card now accounted for 19 percent of impressions, up from nothing in March. The practice was not losing. It was getting pulled into AI answers and showing fewer raw listings as a result. The only owners who panic at a falling total are the ones who never open the surface breakdown.
Why most profiles score under 15 percent
AI surfaces do not guess. They quote facts they can verify, and they pull those facts from structured fields, not from your homepage prose. A profile scores low for boring, fixable reasons. The product and service entries are empty or one line long. The reviews say “great service” instead of “rewired our panel in a day.” The website has no LocalBusiness or Service markup, so the model has nothing clean to read. Google fills the answer with whoever fed it the cleanest facts, and that is rarely the business with the prettiest site.
Here is the honest tradeoff. If you run a plumbing or HVAC shop, summaries are live in your vertical and a low share is costing you booked jobs right now. If you run something Google has not started summarizing, a low share is just noise, and chasing it this month wastes your time. Look at the row first. A share near zero with healthy classic-panel impressions usually means no summaries in your category. A share near zero with strong AI-surface impressions for competitors means you got left out. Those are different problems.
The 90-minute fix: feed the surfaces what they read
If summaries are live in your category and your share is low, the work is concrete and you can finish most of it before lunch. Do it in this order, because each step feeds a specific surface.
First, fill every product and service entry in your profile with a real two-sentence description. The Maps AI card pulls answers straight from these fields when someone asks “does anyone nearby do X.” A blank entry is a silent no. Name the service, the typical turnaround, and the service area. Second, add LocalBusiness and Service schema to the matching pages on your site. This is the structured fact source AI Overviews read before they cite you. A minimal block looks like this:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Plumber",
"name": "Carson Plumbing",
"areaServed": "Boise, ID",
"telephone": "+1-208-555-0142",
"makesOffer": {
"@type": "Offer",
"itemOffered": {
"@type": "Service",
"name": "Emergency leak repair",
"description": "Same-day leak detection and repair, most jobs done in one visit."
}
}
}
</script>
Third, add a short FAQPage block answering the exact questions buyers type: pricing range, response time, service area, warranty. AI Overviews lift these almost verbatim. Fourth, change how you ask for reviews. Stop asking for stars and start asking customers to name the job. A review that says “fixed our AC the same day we called” is a grounding fact the model can cite. “Five stars, great guys” is not. Send the next ten review requests with one prompt: “Could you mention which service we did and how fast.” That single change does more for AI-surface share over a quarter than any homepage redesign.
One warning before you start. Do not stuff keywords into the product descriptions or the schema. The summaries cross-check your structured facts against your reviews and your site copy, and a mismatch gets you dropped, not promoted. If your schema claims “24-hour emergency service” and zero reviews mention an after-hours call, Google treats the claim as unverified and quotes a competitor instead. Keep every field true and specific. The model rewards businesses whose facts agree with each other across the profile, the reviews, and the site, which is exactly the kind of consistency a real operating business produces by default.
What to watch over the next 30 days
Re-open the Performance tab in 30 days and read the AI-surface row again, not the total. Total impressions are sliding for almost everyone as AI answers absorb clicks, so a falling total can sit on top of a rising AI-surface share. That combination is the win: fewer raw views, more of them landing inside the answer customers actually read. If the share moved up and your calls and direction requests held steady, the work paid off, even though the headline number looks worse.
The signal to track this month is how fast Google expands product-grounded summaries into new categories. Maps already answers “who sells X near me” by reading product entries, and that behavior keeps spreading to service verticals. The businesses that filled their entries early will own the answer slot when summaries reach their category, because Google has nothing else clean to quote. That is the whole game now: be the cleanest fact, not the loudest site.
At Atlas Unchained we build the websites and the structured-data plumbing that get local operators cited instead of skipped. If your AI-surface share is under 15 percent and summaries are live in your trade, we can fix the grounding layer in a week. Subscribe to The AU Brief for the one local-marketing move worth making each week, written for operators who own both the site and the budget.
