# AI Local Packs Show 68% Fewer Businesses: What the May 2026 Core Update Means for Multi-Location Brands > Google's May 2026 core update reshaped local search, and AI local packs now surface 68% fewer businesses. What multi-location brands should fix first. - Source: https://socialplaces.io/blog/may-2026-core-update-ai-local-pack-multi-location-brands/ - Author: Conor Young - Published: 2026-06-24 - Tags: Core Update, Local SEO, AI Overviews, AI Search, Google Business Profile, Multi-Location, Reputation --- Google's May 2026 broad core update finished rolling out on 2 June. If your locations lost ground in the local pack this year, the update is a reasonable suspect. It is not the whole story. The bigger shift is what the local pack has become. When Google returns an AI-generated local result instead of the familiar three-pack, roughly two out of every three businesses that would have appeared simply do not. No core update reverses that. ## What the May 2026 core update actually was The update began rolling out on 21 May at 08:40 PDT and Google marked it complete on 2 June at 05:40 PDT, a window of 11 days and 21 hours. There was no companion blog post and no update-specific guidance. The official description was the standard one: a regular update designed to better surface relevant, satisfying content from all types of sites. Two things made it notable. It was the second broad core update of 2026, arriving roughly six weeks after the March update completed on 8 April, a tighter cadence than the three to four month spacing brands had grown used to. And the volatility was heavy, with spikes over the weekends of 23 and 30 May and again in the final 24 hours before completion. Google's own advice is to wait a full week after completion before reading Search Console, then compare that week against a week before 21 May. Anything measured during a rollout is noise. ## The number that matters more than the update Recent industry analysis compared AI-generated local packs against traditional three-packs across the same set of queries. The traditional packs surfaced 18,330 unique businesses. The AI packs surfaced 5,943. That is roughly a third as many, or 68% fewer businesses getting any visibility when the AI format appears. Across 322 markets tracked, the AI version showed fewer businesses than the traditional version in 88% of them. Where the three-pack showed three, the AI pack often shows one or two. Many versions also drop the click-to-call button, which changes the action available, not just the visibility. ![Comparison headed 'The AI local pack shows 68% fewer businesses': for the query 'car wash in Sandton', the traditional three-pack lists your Sandton location first at 4.4 stars, while the AI local pack shows one competitor, fades another, and filters your location out entirely.](./may-2026-core-update-what-changed.png) Two caveats worth keeping honest. The AI local pack still triggers on a minority of local keywords, measured at around 8% and rising, so this is not yet most of your search volume. And the research is US-weighted, so South African brands are reading the direction of travel rather than a current local benchmark. The format is expanding query by query, not flipping on overnight. ## Two selection systems, running side by side The businesses that make an AI local pack are not simply the ones with the most reviews. They are the ones the model can describe with confidence: complete profiles, consistent data across the web, reviews that mention specific services, attributes and staff, and structured pages that are straightforward to read. A separate 2026 study of 1,120 local searches across seven verticals and 13 cities found that 28.5% of businesses appearing in the traditional local pack were absent from AI Mode for the same query. It also found 999 businesses that appeared only in AI Mode and never in the local pack at all. That is the part most brands miss. These are two different systems building two different shortlists. Ranking in one does not put you in the other, and your rank tracker only watches one of them. ## Review response rate is the lever most brands are not pulling Review signals carry roughly 20% of local pack ranking weight in 2026, up from about 16% three years ago, according to the most widely cited practitioner survey in local search. What moves inside that weighting has shifted from raw volume toward velocity, recency and owner response. - Businesses responding to 80% or more of their reviews see measurable local pack gains. - Businesses responding to 90% or more see around 23% more profile views and 18% more direction requests than those responding to fewer than half. - 74% of consumers look for reviews from the last three months. A profile whose newest review is eight weeks old already reads as quiet. - Reviews are also a primary source AI answers draw on when assembling local recommendations, so the same work compounds across both systems. ![Stat card headed 'Review response is doing real ranking work': locations at a 90%-plus response rate see 23% more profile views and 18% more direction requests, with reviews and responses now carrying roughly 20% of local ranking weight, up from about 16% three years ago.](./may-2026-core-update-response-rate.png) The instinct when rankings drop is to push harder on review generation. Volume is the wrong target. A location with 38 reviews from the last two months, all answered, is in better shape than one with 412 reviews where the most recent is from two years ago. ## The franchise gap A single-location business can check its Google Business Profile daily. A brand with 60 or 150 locations cannot, not by hand. The result is variance. Some locations reply within hours. Others have reviews sitting unanswered for six months. Core updates expose that variance, and AI selection punishes it harder, because a location with thin, generic review content gives the model nothing specific to work with. For franchise brands this is not a content quality problem. It is an operational cadence problem, and it is measured per location, not per brand. Your brand average is hiding your worst stores. ![Review response monitor table headed 'Run review response as one queue across every location', showing response rate and average speed per location — Sandton at 94% and 2 hours and Rosebank at 91% and 4 hours on track, Menlyn flagged needs-attention at 67% and 31 hours, Gateway at 88% and 9 hours — with the note that one queue surfaces the location slipping below threshold before its ranking does.](./may-2026-core-update-one-queue.png) ## What to do now - **Audit response rate by location.** Pull the rate for every location, flag anything below 80%, and start there. Do not work off the brand average. - **Respond with specifics.** Name the service, the branch and the team member where you can. Generic thank-you replies add nothing for Google or for an assistant summarising what people say about you. - **Fix profile completeness first.** Missing hours, services, attributes and photos are the cheapest gap to close, and they feed both selection systems. - **Check review recency per location.** A dormant profile loses customers and confidence at the same time. - **Treat Google Posts as engagement, not ranking.** A controlled study across 441 keywords found no local pack movement from Posts. They still support click-through and keep a profile looking active, so keep posting, but do not expect a ranking lift from it. - **Do not chase a review spike.** Google confirmed in May 2026 that its spam policies extend to attempts to manipulate generative AI responses, including AI Overviews and AI Mode. A sudden burst of solicited reviews is a risk, not a shortcut. ## A note on timing Recovery from a core update usually arrives with the next core update rather than in the gap between them. Google's guidance is explicit that there is no specific fix, and that genuine improvements can take months to register. The work you do now is banked for a moment you do not get to choose, which is an argument for starting before the next one lands rather than after. ## How Social Places helps multi-location brands close the gap Everything above is an operational problem before it is a search problem. Both selection systems read the same underlying data, and holding that data accurate and current across 50, 100 or 250 locations is not a task that scales by hand. **[Listings](/listings/) is the foundation.** Store data is managed centrally and synced across Google, Apple Maps, Bing, Facebook, TomTom, Waze and 50+ platforms from one dashboard. Individual and bulk updates, special trading hours loaded ahead of public holidays, data cleansing with manual Street View verification, and Google verification handled directly with Google. LocTech detects and reverses unauthorised user-suggested edits before they spread, Rogue Page Finder locates and suppresses duplicate pages competing with your real listing, and schema markup is applied and audited across every location. That is the structured, consistent data both Google and the assistants read, which is why the completeness gap described above closes here first. **[Store Locator](/store-locator/) turns that data into pages you own.** A branded, SEO-optimised locator plus a landing page for every location, hosted on your own site and synced from Listings, with pages created and closed automatically as locations open and close. Near-me schema, meta and title tags are managed from the dashboard. Each page carries an interactive map, per-location hours and attributes, store-specific events and specials, social links, and booking or online ordering where you run it, with page-level reporting per store. When an assistant needs a current, structured source for one branch, that is the page available to cite. **[Reputation](/reputation/) supplies the trust signals.** Every review, comment and message lands in one inbox with AI-drafted replies in your brand tone, and response time and rating are tracked by location, so the quiet stores surface instead of hiding in the brand average. Updated August 2026: no further broad core update has been confirmed since May. Google's next confirmed ranking changes were the June 2026 spam update and an August 2026 spam update that began rolling out on 18 August. Social Places has also since launched [AI Visibility Tracker](/ai-visibility/), which measures how often ChatGPT and Gemini name each of your locations, benchmarked against a competitor set you choose and split by assistant, query and location. It is built for exactly the gap described above, a visibility problem that shows up in no ranking report. If your locations lost ground this year, the fix is not more reviews. It is a more consistent, more responsive presence on every profile you already own. [Contact Us](/contact/)