/ai-search921 words

AI Overviews appeared on 99% of high-intent local searches in two-month study

A hand-run study of 480 local searches across 12 US markets found AI Overviews on 38% of queries initially and 99% two months later — and in the last round, the AIO fully replaced the map pack.

AI Overviews are taking over the map pack. Here’s what I found
AI Overviews are taking over the map pack. Here’s what I foundAI-generated
  • AIOs appeared on 38% of high-intent local searches in round one and 99% (79 of 80) in the final round about two months later.
  • In the last round, 79 of 80 searches showed no separate map pack; businesses appeared as cards inside the AIO.
  • On 22% of co-occurring searches, the AIO named none of the map pack's three businesses.
  • 373 of 1,023 AIO-recommended businesses (36%) had no website ranking on page one.
  • Conversational searches triggered AIOs first: 66% versus 24% for bare keywords on the first pass.

AI Overviews showed on 99% of high-intent local searches in the final round of a 480-search manual study, up from 38% two months earlier. The tester, a local SEO agency practitioner, ran every query by hand across 12 US markets between September and November. This is tool-level, single-operator data, not a confirmed Google statement. Google has not announced any of the behaviors the study documents.

The study covered 10 service types — home services, legal, healthcare, and restaurants — in three market sizes: big metros (Philadelphia, Phoenix, Nashville), mid-size cities (Worcester, Spokane, Toledo), suburbs (White Plains, Naperville, Cedar Park), and small towns (Thomasville, Marshall, Cottonwood). Each service was searched two ways, bare ("emergency plumber [city]") and conversational ("need an emergency plumber [city]"), in incognito windows, signed out, twice per market.

How fast did AI Overviews spread?

In the first eight markets, AIO appearance rose from 38% to 66% in about two weeks. In the last four, it went from 60% to 99% in about four weeks. Across all 480 searches, Google served an AIO on 292 (61%).

Worcester, Massachusetts, showed the sharpest change: one of 20 searches carried an AIO on the first pass, 12 of 20 on the second. Philadelphia went from 4 to 19. Ten of the 12 markets gained AIOs between rounds. City size did not predict coverage; Phoenix, Cedar Park, and Cottonwood all hit 100% while spanning metro, suburb, and small-town categories.

AIOs also persisted once they appeared. Of the 108 AIOs in the first pass, 106 remained on the second. Every service category grew, with dentists tripling from 5 of 24 searches to 16 of 24.

The takeaway in plain words: AIO coverage expanded everywhere the tester looked, and it rarely retreated.

Did the AI Overview replace the map pack?

In the final round, 79 of 80 searches showed no separate map pack at all. Local businesses appeared as cards inside the AIO instead, each carrying name, rating, review count, and call, directions, and website buttons. Four weeks earlier, those same four markets had shown an AIO above a regular map pack on all 48 AIO searches.

The cards were not simply the old pack relocated. The AIO named about five businesses per search against the pack's three. Of 236 first-pass pack businesses, 154 (65%) made it into the cards. On 9 of 79 searches, none did. Nashville kept the fewest, at 42%.

The layout also admitted smaller competitors. One in five cards (81 of 379) belonged to a business with fewer than 50 reviews. On a Cottonwood, Arizona, plumber search, the first card went to a plumber with one review.

The tester captured this round on a single day. It could be a test or a partial rollout. That caveat matters; a one-day snapshot cannot confirm a permanent layout change.

Did the AIO name the same businesses as the pack?

No, frequently not. On the 145 searches where an AIO sat above a map pack, it omitted at least one pack business 114 times (79%). On 32 searches (22%), it named none of the pack's three businesses. It named all three on only 31 searches (21%).

The opening sentence carried the strongest divergence. In 26 of 70 tracked searches (37%), the opening link pointed to no pack business at all. On a Phoenix dentist query, it linked Downtown Smiles, a non-pack practice, before any pack dentist appeared lower down.

Takeaway: map pack rank did not guarantee AIO inclusion, and pack exclusion did not bar it.

Where did AIO picks come from?

More than a third of recommended businesses (373 of 1,023) had no website on page one. That held at 36% even after the AIO replaced the pack. After business websites, the AIO most often cited Yelp, then industry directories such as Justia for lawyers and Zocdoc for healthcare, plus Facebook and Reddit.

On one Cedar Park, Texas, roofing query, every named business came from a single "5 best roofing companies in Cedar Park" blog post — and none of the four roofers were based in Cedar Park. On the second pass, the same search named five different roofers. One list can place a business in the answer but cannot keep it there.

Reviews alone did not decide inclusion. On 46 of 144 searches (32%), the AIO named a sub-50-review business outside the pack while the pack showed one with 200 or more. In White Plains, New York, a personal injury AIO skipped three pack firms holding 180 to 229 reviews and recommended four firms with 6 to 47 each.

What triggered AIOs first?

Conversational phrasing did. First pass: 66% of conversational searches got an AIO against 24% of bare ones. By the second pass, bare searches reached 66% and conversational ones 88%. Estate planning showed the pattern in all 12 markets — 11 of 12 conversational searches carried an AIO on pass one, versus 2 of 12 bare. Conversational AIOs also gave full independent answers more often: 83% versus 36% for bare queries.

The author's practical guidance follows the data: keep Google Business Profiles accurate, since the cards pull GBP names, ratings, and addresses; build a page for every city served; pursue third-party "best of" lists; and publish FAQ content in customer phrasing with LocalBusiness schema.

These are one researcher's field observations. The study lacks stated confidence intervals and replication. Monitor whether the no-map-pack layout spreads beyond those four markets and whether GBP data continues to drive card selection.

via Semrush Blog (Source)

More from Elena Vasquez

Elena Vasquez

Show full bio

Correspondent covering people, launches and day-to-day practice in Search rankings and algorithm updates, generative engine optimization at SERP Journal.

227 articles