No. 028
One test says 93 of 100 brands. The study citing it says 63 of 100.
Ahrefs' branded AI Overviews study cites a 93-of-100 brand test and reports 63 of 100 from its own panel. Five published numbers, five denominators.
93 of 100vs63 of 100
What was published, and when
| Date | Who | What they published |
|---|---|---|
| 2026-09-30 | Chris Long, co-founder of Nectiv, on LinkedIn | 93 of 100 enterprise brands had an AI Overview in one SerpApi scrape that morning |
| 2026-09-30 | Search Engine Roundtable | Covered Long’s post, checked “several” brands by hand, noted Google’s own name and news publishers showed none |
| 2026-10-01 | DemandSphere | A daily series for September, 26.12% on the 1st to 82.06% on the 29th, headlined “tripled” |
| 2026-10-01 | Search Engine Land | Covered the DemandSphere series as a jump “from 26% to 80%“ |
| 2026-10-01 | Inside AI | ”Google AI Overviews Hit 93% of Branded Searches, Tripling in September” |
| 2026-10-02 | Ahrefs | 232.8 million US desktop crawls, 82.91% on 29 September, a 73.3% September average, and its own 63 of 100 panel |
Six publications in three days, and by the end of it the number in circulation was a percentage that nobody had measured.
Five numbers, five denominators
| Number | Source | What was actually counted | Scope | Period |
|---|---|---|---|---|
| 93 of 100 | Chris Long, 30 Sep | Brand-name queries with an AI Overview present, one SerpApi scrape | Location and device not stated | One morning |
| 63 of 100 | Ahrefs, 2 Oct | Exact-name queries for 100 Interbrand brands, 14 weekly checkpoints | US desktop | Reading of 30 Sep |
| 82.91% | Ahrefs, 2 Oct | Share of search-result crawls for branded keywords | US desktop | Single day, 29 Sep |
| 73.3% | Ahrefs, 2 Oct | Share of branded keywords triggering an AI Overview | US desktop | Monthly average, September |
| 26.12% to 82.06% | DemandSphere, 1 Oct | Share of tracked branded keywords returning an AI Overview, once per day per keyword | All markets, all devices | 1 to 29 Sep, daily |
The 83% in Ahrefs’ own title is the 82.91% rounded up. That is one day, on one device class, in one country, expressed as a share of crawls. The 73.3% in the same article is a monthly average of keywords. Those two figures appear in the same paragraph and differ by nine and a half points, because they answer different questions. A reader who takes “83%” as the state of play has taken the highest single day of the series as the level.
DemandSphere’s number is the honest outlier, and the reason is in its method note:
Branded keywords tracked daily through DemandMetrics, measured once per day per keyword, all markets, all devices. We’re working on breakdowns across devices and major markets. An AI Overview is counted as present when the SERP returns an AIO for the query, regardless of whether brand itself is cited in the AIO.
All markets and all devices, against Ahrefs’ US desktop. The number of tracked keywords is not stated anywhere on the page. Both companies sell rank-tracking products, which is not a reason to discount either dataset, but it is a reason to read the scope line before the headline.
The same endpoint, a different path
Here is the part of the story that the word “tripled” hides. This is DemandSphere’s full published series, read from its own page:
| Date | Presence | Date | Presence | Date | Presence |
|---|---|---|---|---|---|
| 09-01 | 26.12% | 09-11 | 33.38% | 09-21 | 28.59% |
| 09-02 | 27.21% | 09-12 | 33.28% | 09-22 | 31.28% |
| 09-03 | 27.25% | 09-13 | 34.27% | 09-23 | 56.97% |
| 09-04 | 26.56% | 09-14 | 28.57% | 09-24 | 55.81% |
| 09-05 | 27.20% | 09-15 | 30.10% | 09-25 | 29.98% |
| 09-06 | 35.45% | 09-16 | 28.57% | 09-26 | 69.21% |
| 09-07 | 34.87% | 09-17 | 30.64% | 09-27 | 90.48% |
| 09-08 | 33.78% | 09-18 | 57.43% | 09-28 | 80.23% |
| 09-09 | 32.82% | 09-19 | 31.15% | 09-29 | 82.06% |
| 09-10 | 28.58% | 09-20 | 46.62% |
Four days before the late-September surge already broke the band: 57.43% on the 18th, 46.62% on the 20th, 56.97% on the 23rd, 55.81% on the 24th. Each is followed by a day back in the low thirties. The 18th reads 57.43% between a 30.64% and a 31.15%.
A series that jumps twenty-six points and falls back the next day, four separate times, is not obviously a rollout. It is at least as consistent with test buckets or with sampling, and DemandSphere does not claim otherwise. What it does say is this, and it is the most important sentence on the page: “Google has not made any announcements on this change.”
Three arithmetic notes, which are ours and not either company’s figures. The mean of DemandSphere’s 29 daily values is 40.64%. The mean of the 25 days before the surge is 34.26%. And 82.06 divided by 26.12 is 3.14, which is where “tripled” comes from: it is the last day against the first day, not the month against a previous month.
That last point is worth sitting with. Ahrefs’ September average for branded keywords is 73.3%, up from 65.1% in August and 61.3% in July. On Ahrefs’ own series, September is a twelve-point move on a base that was already above sixty. On DemandSphere’s series, September starts at twenty-six. Two trackers, one month, and a thirty-three-point disagreement about where the month began.
Where the AI Overview sat on the page
Both of the panel results carry a second measurement that travelled less well than the first.
Ahrefs, on its 63:
More importantly, of the 63 AI Overviews in our sample, 48 appeared in position one (76.2%), with an average position of 1.49.
Search Engine Roundtable, looking at Long’s brands by hand the same day:
And most of the examples have the AI Overview showing up lower down in the search results page, not at the top of the page.
These are not necessarily in conflict, because they are different brand sets on different days, and Roundtable’s is an eyeball check of a handful. But they are the difference between a feature that displaces the first organic result and a feature that sits below the fold. “Present” is one measurement. “Present at position one” is another. Only one of the five headline numbers in this story is about position at all.
How 93 of 100 brands became 93% of searches
Long’s own post carried both framings. The body is precise: “I ran a quick scrape with SerpApi this morning and found that 93 out of 100 enterprise brands had AI Overviews present for them.” The headline above it is not: “Google is showing AI Overviews ON THE MAJORITY OF BRAND QUERIES!!!! 93% of the brands I tested have an AI Overview”.
Ninety-three per cent of one hundred tested brands is ninety-three brands. It is not ninety-three per cent of branded queries, and the hundred brands were not a sample of queries. By 1 October, Inside AI had printed exactly that conversion:
Google’s AI Overviews now appear on 93% of branded search queries, according to independent testing by SEO analyst Chris Long.
The same article added a backstory: “The company paused the feature in August following the Hugging Face breach and mounting public criticism”, and dated the broad launch of AI Overviews to May 2026. Neither claim appears in Long’s post, in Search Engine Roundtable’s coverage, in DemandSphere’s post, in Search Engine Land’s write-up, or in Ahrefs’ study. We swept all five for the words “pause”, “paused” and “Hugging” and found nothing. On the launch date, Search Engine Journal put the US default rollout in May 2024, two years earlier than Inside AI’s date.
TechJuice ran the same two additions, nearly word for word, on the same day, and added a causal claim of its own: that the pattern suggests “Google accelerated rollout following internal decisions about AI prioritization”. DemandSphere’s page says Google has announced nothing. Neither of those two articles links to any of the sources it describes.
This is the actual failure mode in AI-search coverage right now. The underlying research was careful and labelled. Two retellings turned a count into a rate, invented a mechanism for it, and published faster than the study they were summarising.
What none of it measures
Clicks. Not one of the five numbers is a click, a session or a referral. Search Engine Land said so plainly in its own piece: the impact on branded click-through “remains unknown”, because neither Long’s test nor DemandSphere’s data measured clicks.
Whether the brand is in the answer. DemandSphere counts an AI Overview as present whether or not the brand is cited in it. For a brand-name query, that distinction is most of the commercial question.
Anything after 29 or 30 September. Both datasets stop there. On Long’s post, Lily Ray replied: “I’m not seeing this today. And now it looks like when they do show AIO, it’s under both the organic and paid listing. Maybe they turned it off already?” That is one practitioner’s observation and not a measurement, which is exactly why it belongs in this section rather than in the tables above.
Cause. No source in this story has an explanation, and the one company with a daily series says Google has announced nothing.
If you want to check your own brand queries
Write down the denominator before the number. For any AI Overview figure you are handed, four facts decide whether it is comparable to yours: country, device, whether it counts crawls or keywords, and whether a single day or an average. Ahrefs’ 82.91% and 73.3% differ on the last one alone, in the same article.
Measure presence and position separately. Forty-eight of Ahrefs’ sixty-three were at position one. Roundtable saw most of its examples lower down. A dashboard that records only presence cannot tell you which of those two things happened to you.
Count your own brands, not a percentage. The most defensible number in this entire story is a count: 63 of 100, on a named list, on a stated date, by a stated method. If you track twenty brand queries daily and write down how many return an AI Overview, you will have a cleaner series for your own brand than any of the five figures above, and you will know its denominator.
Do not use “93%” at all. It does not exist as a measurement. The underlying fact is 93 of 100 enterprise brands in one scrape on 30 September, and that is a sentence you can defend.
Sources
- Ahrefs, “AI Overviews Currently Appear for 83% of Branded Searches”, 2 October 2026: ahrefs.com/blog/ai-overviews-on-branded-searches
- DemandSphere, “AI Overviews on branded queries tripled in September”, 1 October 2026: demandsphere.com/blog/branded-ai-overviews-september-2026
- Search Engine Roundtable, “Google AI Overviews Return For Many Large Brand Names”, 30 September 2026: seroundtable.com/google-ai-overviews-large-brand-names-42195.html
- Inside AI, “Google AI Overviews Hit 93% of Branded Searches, Tripling in September”, 1 October 2026: insideai.news
- TechJuice, “93% of Brands Now Displaced by Google AI Overviews Search”, 1 October 2026: techjuice.pk
- Search Engine Journal, on the May 2024 US default rollout, 11 September 2026: searchenginejournal.com
Chris Long’s post and Lily Ray’s reply are quoted from the public LinkedIn thread dated 30 September 2026. Search Engine Land’s article of 1 October 2026 is quoted from the copy read during research; the page refuses automated requests, so it is not linked here as a verifiable source.