A Publisher Put AI Overviews in an SEC Filing. Revenue Fell 1%.

People Incorporated's Q2 8-K blames Google AI Overviews for a 22% Core Sessions decline. The same filing shows session-based revenue fell just 1%.

2026-08-09· 5 MIN

The short answer

Publishers have spent two years telling reporters that AI Overviews are eating their traffic. Almost none of them put a number on it in a document where being wrong has consequences.

On 3 August 2026, People Incorporated did. Its Q2 2026 earnings release, filed as an exhibit to a Form 8-K, attributes a 22% decline in Core Sessions “due primarily to the impact of the growing prominence of Google AI Overviews on Google search results.”

That is the strongest single piece of publisher-side evidence on this question we have seen, because of where it appears rather than what it says.

Then there is the second number, four lines down the same table. Digital session-based revenue — the revenue that depends on those sessions — fell 1%.

Both facts come from one filing. A post that cites only the first one is telling half the story the company itself filed.

Where AI Overviews actually appears in the filing

It appears once. In a subordinate clause. Explaining an advertising line item.

The full sentence, verbatim:

Higher open programmatic advertising revenue due to higher rates, partially offset by lower impression volumes driven by a 22% decline in Core Sessions, due primarily to the impact of the growing prominence of Google AI Overviews on Google search results

The exhibit sits in the filing’s document index if you want to read the whole thing rather than take our word for any of this.

This is worth being precise about, because the framing will get inflated. People Incorporated did not publish a headline finding about AI Overviews. It did not quantify how much of the 22% is AI Overviews as opposed to everything else bundled into “primarily.” It named a cause, once, in passing, while explaining why impression volumes fell.

What makes it evidence is the venue. An earnings-call answer is a person talking. An 8-K exhibit is a filed document, and the company’s own lawyers read it first.

The number that moved

The filing reports two session metrics, and defines both.

Metric	Q2 2026	Q2 2025	Change
Total Sessions (millions)	1,891	2,444	-23%
Core Sessions (millions)	1,707	2,202	-22%

Total Sessions is defined in the filing as “unique visits to all sites that are part of People Inc.’s network.” Core Sessions is “a subset of Total Sessions that comprises unique visits to People Inc.’s most significant” properties.

Roughly half a billion sessions gone year over year. Whatever share of that AI Overviews owns, the direction is not ambiguous and the company is not hedging about it.

The number that didn’t

Here is the same filing’s revenue table for the digital segment.

Line	Q2 2026	Q2 2025	Change
Digital session-based revenue	$164.6M	$165.6M	-1%
Digital non-session-based revenue	$125.4M	$108.4M	+16%
Total digital revenue	$289.9M	$274.0M	+6%

Session-based revenue is defined in the filing as “revenue related to advertisements served or performance marketing referrals initiated during a session.” It is the line most directly exposed to a session collapse. Sessions fell 22-23%. That revenue fell 1%.

Total digital revenue went up 6%.

There are two honest readings and they are not the same.

The first is that per-session monetisation improved enough to absorb almost the whole decline — the filing attributes the flat advertising line to “higher rates” on open programmatic. Fewer sessions, worth more each.

The second is that the business changed shape. Non-session-based revenue grew 16% and now supplies $125.4M of a $289.9M digital segment. That growth is attributed in the filing to direct-sold advertising, D/Cipher+, and a Feedfeed acquisition — none of which depend on Google sending anyone anywhere.

Neither reading supports “AI Overviews did not hurt this publisher.” Both undercut the simpler claim that traffic loss converts directly into revenue loss. On this quarter’s evidence, for this publisher, it did not.

Why one quarter is not a trend

The obvious objection to everything above: this is a single quarter from a single company, and the company has an interest in how the story lands.

That objection is correct, and it cuts in both directions. The 22% figure is as much a single-quarter datapoint as the 1% figure is. Anyone quoting the session decline as proof of a general collapse in publisher economics is making exactly the error they would object to if the number ran the other way.

What the filing supports is narrow and worth stating plainly. One large publisher, in one quarter, told the SEC that AI Overviews was the primary driver of a 22% decline in its most significant properties’ sessions, and in the same document reported that the revenue line most exposed to that decline moved by 1%.

What it does not support: any figure for how far Google search traffic specifically fell. The company disclosed none, and the approximations available from the earnings call do not reconcile cleanly with the published figures. We are not going to derive one and neither should anyone else.

What we deliberately did not use

This filing is going to be quoted a lot over the next week, and some of what gets attached to it will not be in it.

The same 8-K notes that People Incorporated ceased operations of its Search business and that a Google Services Agreement expired on 30 April 2026. That refers to a discontinued product segment of its own. It is not a description of lost Google referral traffic, and the two will get conflated because they share vocabulary.

We also left out every crawl-to-referral ratio currently circulating in secondary aggregator posts. We tried to verify them and found mutually inconsistent versions of the same supposed figures across sources. Until they trace to a dated primary publication, they are not numbers.

What would change this reading

A second quarter. If Q3 2026 shows session-based revenue finally following sessions downward, the 1% becomes a lag rather than a decoupling, and this post’s second half is wrong.

The reverse test matters too: if non-session revenue keeps growing fast enough to carry the segment, then the interesting story about AI Overviews for large publishers is not traffic. It is that traffic stopped being the business, and the filings are where that shows up first.