Local Rankings Held. Calls Fell 11.9%. Google Maps Kept the Customer.

Local rankings have stopped predicting local revenue. Across US Google Business Profiles tracked by GMBapi, calls fell 11.9% and website clicks fell 12.5% year over year in Q2 2026 — while direction requests rose 21.1%. The demand did not disappear; it stopped passing through your website and your phone line on its way to your front door.

That single reversal breaks the assumption most local reporting is built on: that a stable map-pack position should produce stable calls. It no longer does, and no amount of rank tracking will show you why.

Key takeaways

• US Q2 2026 vs Q2 2025 (GMBapi): calls −11.9%, website clicks −12.5%, direction requests +21.1%.
• Mobile and desktop split in opposite directions. Mobile Search impressions fell 20.1% while desktop Search impressions rose 13.9% — the same profile, two different stories.
• AI local packs are real but still rare. Sterling Sky found them on roughly 7% of tracked keywords, so they cannot be the sole cause of a double-digit call decline.
• When AI packs do appear, the shelf is smaller. They surfaced 5,943 unique businesses against 18,330 in traditional 3-packs — about 32% as many — and in 88% of 322 markets analyzed the number of visible businesses dropped.
• AI assistants recommend almost nobody. SOCi's 2026 Local Visibility Index put ChatGPT recommendation at 1.2% of locations versus 35.9% average appearance in Google's local 3-pack, across nearly 350,000 locations.

Why are calls and clicks falling while rankings hold steady?

Because ranking and outcome measure two different things, and they have quietly decoupled.

A map-pack position measures where you sit in a list. A call or a website click measures a customer leaving that list to go somewhere else. Google has spent two years removing reasons to leave: hours, menus, photos, booking, review summaries and Q&A now resolve inside the profile itself. The customer who once called to ask "are you open until seven?" reads it, taps directions, and drives over.

Under that behavior, a flat ranking with falling calls is not a failure. It is the same demand exiting through a door your analytics never instrumented.

The tell is direction requests. They are the only major Google Business Profile action that rose in the US in Q2 2026, and they are the action closest to an actual visit. A profile losing calls while gaining directions is not losing customers — it is converting them one step earlier, out of sight of your website.

What does the Q2 2026 data actually show?

Source: GMBapi aggregate Google Business Profile data, as reported by Search Engine Land.

Read the impression rows before the action rows. Mobile Search fell 20.1% while mobile Maps rose 30.4% — the same searcher, relocated. Discovery moved from the Search results page into the Maps surface, and the Maps surface is built to end the journey rather than hand it off.

Desktop tells the opposite story: impressions up on both surfaces. Desktop is where website clicks still happen, which is precisely why website clicks fell less catastrophically than the mobile experience alone would predict.

One caution on this dataset. These are aggregate figures from one vendor's profile panel, reported by a trade publication; GMBapi's own write-up is not publicly readable at the time of writing. The direction of travel is corroborated by independent local-search analysts, but treat the decimal places as indicative, not audited. Regional figures for the EU and UK circulate alongside these, including a swing in UK mobile Maps impressions dramatic enough to suggest a panel composition change rather than a market change — we have left those out here because we could not verify them against a primary source.

Are AI local packs the cause?

Partly, and less than the headlines imply. This is the tension most coverage skips.

Sterling Sky's tracking found AI local packs appearing on about 7% of tracked keywords. A feature present on one query in fourteen cannot, by itself, remove 12% of a country's calls. Anyone selling you an AI-pack panic is overreaching.

What AI packs do is change the shape of the result when they appear, and the change is severe:

Source: Sterling Sky, The State of Local SEO in 2026.

The missing call button matters more than the missing third slot. On mobile, the phone icon is the conversion. Remove it and a query that would have produced a call produces a tap into the profile instead — where the customer finishes without ever touching your site. Sterling Sky's profile-level analysis, drawn from 179 Google Business Profiles across 34 US law firms, showed mobile click-to-call falling while desktop website clicks held roughly steady. That asymmetry is a fingerprint: it points at an interface change on mobile, not at a demand collapse.

So AI packs are a genuine, growing pressure on a small slice of queries — not the explanation for the whole curve.

What else is absorbing the clicks?

Paid inventory, and it grew faster in 2025 than AI packs did.

Sterling Sky's tracking recorded ads inside the local pack rising from 1% of reports in early 2025 to almost 22% by December 2025, and Local Services Ads from roughly 11% of queries in early 2025 to 31% by November 2025.

That is a structural change to the top of the local result, and it is the boring explanation nobody is writing about. Nearly a third of tracked local queries now carry Local Services Ads above the organic pack. Every one of those pushes the free listing down a screen — on mobile, often below the fold entirely. A business whose rank held perfectly still in 2025 was nonetheless demoted in pixels.

Three forces are compressing organic local clicks at once: in-Maps journey completion, expanding paid inventory, and (on a minority of queries) AI packs with fewer slots and no call button. Attributing the whole decline to any one of them will lead you to the wrong fix.

How selective is AI recommendation for local businesses?

Brutally selective — and inconsistently accurate.

Source: SOCi 2026 Local Visibility Index, nearly 350,000 locations across 2,751 multi-location brands.

SOCi framed the gap as AI visibility being three to thirty times harder to achieve than ranking well in traditional local search. In retail, only 45% of the top 20 brands by traditional local visibility also appeared among the top 20 most recommended by AI platforms — the winners are not the same companies.

Two details deserve more attention than they have received.

Accuracy is not a rounding error. Roughly a third of what ChatGPT and Perplexity said about business profiles was wrong, against 100% for Gemini, which grounds directly in Google Maps. If a third of your hours, addresses or service descriptions are misstated by an assistant, keeping your Google Business Profile flawless is doing double duty: it is your ranking asset and the ground truth that the most accurate assistant reads from.

Higher selectivity comes with a higher review bar — but not uniformly. Locations recommended by ChatGPT averaged 4.3 stars; Gemini's averaged 3.9. The most selective assistant demanded the best ratings, while the Maps-grounded one was willing to recommend more ordinary businesses. Review quality is a stronger lever on the assistants that recommend the fewest businesses, which is exactly where a marginal gain is hardest and most valuable.

Which metrics should replace rank tracking?

Rank is now a leading indicator of nothing in particular. Replace it as your headline number, not as a diagnostic.

Two practical notes. First, most rank trackers still do not report AI local packs, so a report showing "no change" may simply be blind to the feature that changed. Second, if you compare quarters, compare like for like — year over year, not against the previous quarter, since local demand is seasonal and a quarter-on-quarter drop can be a calendar artifact dressed up as a crisis.

If you are rebuilding a reporting template around this, our guide to choosing AI-search KPIs walks through the same shift for non-local businesses, where the equivalent lost metric is the click that never happens after an AI Overview.

What should a local business change this quarter?

1. Rebaseline your reporting on directions and calls, year over year. If directions are rising while calls fall, you are converting, not declining. Say so in the report before someone panics at the click chart.
2. Fix the mobile path first. The declines are concentrated where the phone icon lives. Make sure the number on your profile is correct, answered, and not routed to a menu tree.
3. Treat your Google Business Profile as the canonical record. Gemini's 100% accuracy comes from reading Maps directly. Everything wrong on your profile is wrong in the most accurate assistant too.
4. Work review velocity, not just the average. Assistants that recommend 1–7% of locations skew toward highly rated ones. A 4.1 that is climbing beats a 4.4 that stopped moving two years ago.
5. Measure how far down you actually are. Track whether your queries return Local Services Ads or an AI pack. Pixel position, not list position, is what your customer sees.
6. Give assistants something to quote. Structured hours, services, and clear on-page answers are what get extracted. The same discipline that earns an AI citation earns a recommendation.

Frequently asked questions

Why are my Google Business Profile calls down if my ranking has not changed?

Because ranking measures list position and calls measure someone leaving the list. Google now resolves hours, photos, reviews and directions inside the profile, so customers complete the journey in Maps. US calls fell 11.9% year over year in Q2 2026 while direction requests rose 21.1% — the demand moved, it did not vanish.

Are AI local packs responsible for the drop in local calls?

Only partly. Sterling Sky found AI local packs on roughly 7% of tracked keywords, too few to explain a double-digit national decline. Where they do appear they show 1–2 businesses instead of 3, surface about 32% as many unique businesses, and often omit the click-to-call button.

What is the best replacement metric for map-pack rank?

Direction requests, tracked year over year and split by mobile versus desktop. Directions are the action closest to a physical visit and the only major profile action that grew in the US in Q2 2026.

How often do AI assistants recommend a given local business?

Rarely. SOCi's 2026 Local Visibility Index measured recommendation rates of 1.2% on ChatGPT, 7.4% on Perplexity and 11% on Gemini, against a 35.9% average appearance rate in Google's local 3-pack, across nearly 350,000 locations.

Do reviews affect whether AI recommends my business?

They correlate strongly. Locations recommended by ChatGPT averaged 4.3 stars versus 3.9 for Gemini, so the more selective the assistant, the higher the review bar. Velocity and recency matter alongside the average, because assistants weigh what looks current.

Should I stop tracking local rankings entirely?

No — keep rank as a diagnostic, drop it as the headline KPI. It still tells you whether Google understands your relevance and proximity; it no longer tells you whether the phone will ring.

Where to start

If your calls are falling and you do not yet know whether it is Maps, ads, or AI packs taking them, start by confirming what machines can actually read on your site. Run the crawler checker to see which AI crawlers you allow, and the meta tags checker to confirm your pages describe your business the way you would want an assistant to repeat it. Both are free and take under a minute.

For the wider picture behind this shift — why impressions can rise while clicks fall across every kind of site, not just local ones — read AI Search Visibility: Why Clicks Are Falling and Brand Demand Isn't.