Key findings
- 1.52.4% of the 317 US commercial keywords we measured returned an AI Overview between 31 July 2026 and 19 September 2026.
- 2.AI Overview presence ranged from 0% in home services to 96.7% in b2b accounting software — a 96.7-point spread across 11 industries.
- 3.Not one of the 23 "near me" or "local" keywords in the sample returned an AI Overview, against 56.5% of the other 294.
- 4.Only 19 of 317 keywords (6%) showed an AI Overview and a local pack on the same SERP.
- 5.95.7% of those local-intent keywords returned a local pack instead.
- 6.164 of the 166 AI Overviews we recorded (98.8%) linked out to at least one source.
- 7.Question-phrased keywords returned an AI Overview 56% of the time against 52.1% for the rest — a gap our sample of 25 is too small to call meaningful.
Industry is a proxy. Intent is the mechanism.
The headline number — 52.4% of 317 keywords — is the least interesting thing in this dataset, because almost nobody competes across all 11 industries at once. What matters is the spread. B2B accounting software keywords returned an AI Overview 96.7% of the time. Home services keywords returned one 0% of the time. Those are not the same search engine, and a strategy built on the average is built for a SERP neither of them sees.
Read down the list and the pattern is not really about industry at all. The three software categories at the top share one property: the query has no place attached to it. The two at the bottom — home services and legal services — are queries where the searcher wants somebody nearby. Google appears to be making a judgement about what kind of answer the query deserves, and a summarised paragraph is not its answer to "who can come and fix this today".
Local intent and AI Overviews barely coexist
Of the 317 keywords with an observed SERP, 19 (6%) showed both an AI Overview and a local pack. Isolate the keywords carrying explicit local intent — the ones containing "near me" or "local" — and the effect is absolute: 0 of 23 returned an AI Overview, while 95.7% returned a local pack. Every other keyword in the sample returned an AI Overview 56.5% of the time.
This is a small sample of local-intent phrasings (23 keywords), so the honest reading is "we did not observe one", not "it cannot happen". But the direction is consistent with the industry-level numbers, where the two verticals dominated by local packs are also the two with almost no AI Overviews at all. For a business competing on "near me" searches, the AI Overview is — for now — somebody else's problem.
Where the click is at risk, and where it isn't
Our CTR study measured what an AI Overview costs a B2B page once it appears. This study answers the prior question: whether it appears at all. Put together, they say the exposure is concentrated. A B2B software company publishing definitional and comparison content is operating in a category where an AI Overview is close to guaranteed —96.7% in b2b accounting software — and where the question of what gets cited is the whole game. A home services business is not in that fight yet.
One number is worth holding onto in both cases: 164 of the 166 AI Overviews we recorded (98.8%) linked out to at least one source. Where the AI Overview shows up, it is still handing out links — to whoever it decided to cite.
Methodology
Sample. 11 industries, each defined by a single seed term. For each seed we took the top 30 US keywords by monthly search volume that contain the seed as a phrase, with a floor of 500 monthly searches to exclude noise — 330 keywords in total.
Data source. SERP feature composition from the Ahrefs Keywords Explorer API (serp_features), country US. A keyword counts as showing an AI Overview when its feature list contains ai_overview, and as showing a local pack when it contains local_pack.
Date range. Each keyword carries the date its SERP was last crawled. The observations here span 31 July 2026 to 19 September 2026; they are not a single snapshot of one moment, and AI Overview presence is known to fluctuate.
Exclusions. 13 of the 330 keywords had no crawled SERP on record. They are excluded from every denominator rather than counted as "no AI Overview" — the absence of an observation is ours, not Google's. All percentages here use n=317.
Definitions. "Local intent" means the keyword text contains "near me" or "local"; this is a lexical test, not a judgement about the searcher. "Question-phrased" means the keyword begins with what, how, why, when, which, who, is, are, do, does or can.
Limitations. Ahrefs' crawl is a sample of one location and device profile per keyword, so personalised and mobile SERPs may differ. Per-industry cells are 25–30 keywords, which is enough to separate 0% from 97% and not enough to rank two neighbouring verticals against each other. Seed-term matching means each vertical is a slice of a market, not the market. And presence is not impact: this study says where AI Overviews appear, not what they cost — that is the CTR study's question.
Reproducing it. The aggregated per-industry data is published as a CSV download. The raw per-keyword pulls and the analysis script are in the repository that builds this site.
Per-industry counts and rates for all 317 measured keywords, as CSV. Free to use with attribution.
Download the CSVHow to cite this
Using a figure from this page? Please credit it like this:
Source: Indexed, "Where AI Overviews Actually Appear: 11 Industries, 317 Keywords", September 2026, joinindexed.com/studies/ai-overview-by-industry-2026
Updates to this study
- First published. Q3 2026 baseline: 330 keywords collected across 11 industries, 317 with an observed SERP.
We run this analysis against your own keyword set — which terms trigger an AI Overview, who gets cited on them, and what that is costing you.
Talk to a strategist