Google Showed Ziff Davis Links 2.7 Billion Times in AI Answers Last Quarter. It Sent About a Tenth as Many Clicks.
For two years publishers have argued that AI answers are built on their work and pay them almost nothing for it. The argument usually stalls on evidence: the AI side of the ledger is hard to see. Last week one of the largest digital publishers in the US put both sides of the ledger in the same sentence, using a number Google itself supplies.
On its third quarter 2026 earnings call on 6 November, Ziff Davis, the owner of CNET, IGN, PCMag, Everyday Health and RetailMeNot, said organic Google referrals to its properties fell about 37% year on year. In the same quarter, Google's Search Console reported that links to Ziff Davis content were shown inside AI Overviews and AI Mode roughly 2.7 billion times, nearly ten times the clicks those sites received from Google Search (named-party figures, Ziff Davis CEO Vivek Shah).
That is the supply side of AI advertising stated as plainly as a public company is ever likely to state it.
What did Ziff Davis actually report?
The headline numbers come from the company's 8-K filing of 5 November and the call the next morning.
Revenue from continuing operations was $289.3 million, down 5.6% from $306.5 million a year earlier (reported, Ziff Davis Q3 2026 filing). Advertising and performance marketing revenue fell 7.7%, while subscription and licensing revenue fell 2.7% (named-party figures, CFO Bret Richter). The Technology & Shopping segment, home to CNET and the brands most exposed to search, fell 3.8% and booked a $29.6 million goodwill impairment in the quarter (reported, company filing).
On search, Shah gave three numbers. Organic Google referrals fell about 37% in Q3 2026. They had already fallen about 23% in Q3 2025. Together that is a two-year decline of more than 50% (all named-party figures, Ziff Davis). He told analysts the company is not planning for a search recovery and expects referrals to keep falling.
What is an "AI impression", and why does it matter?
An AI impression is Google's count of how many times a link to a page was shown inside one of its generative AI features, such as AI Overviews or AI Mode. Google added it to Search Console reporting this year. It is not a click and it is not proof that a person read the page. It is a count of the moments Google's AI used a publisher's content as a source and displayed the link next to its answer.
That makes it the first platform-supplied metric that measures the read rather than the referral. For most of the web's history those two moved together: if Google surfaced your page, a share of people clicked, and the click carried an ad impression. The Ziff Davis numbers show the two have separated. Roughly 2.7 billion AI impressions against clicks on the order of 270 million is a ratio of about ten to one (blankspace estimate, derived from the company's "nearly 10 times" comparison).
A caution on reading it. Impressions and clicks are different units, and an AI answer can cite several sources at once. The ratio is directional, not an exchange rate. But the direction is the point.
Why does a 37% traffic drop only cost 5.6% of revenue?
An analyst asked exactly this. Shah's answer was that most of Ziff Davis does not live on Google traffic: subscriptions, apps, newsletters, Humble Bundle, IPVanish. Earlier in the year the company said roughly a third of revenue depends on web traffic and that search is around 40% of that traffic (named-party framework, Ziff Davis, not updated this quarter). That puts search-exposed revenue at something like 13% of the total (blankspace estimate).
That diversification is a luxury. Ziff Davis has $1.5 billion of cash and investments and four segments. A single-title publisher whose revenue is mostly display advertising against search traffic does not have a RetailMeNot app to fall back on. For that publisher, the 37% is closer to the whole story.
What is Ziff Davis doing about it?
Three things, by its own account. It is moving monetisation to surfaces Google does not control: vertical video, social commerce, apps and direct audiences. It is cutting cost, with headcount expected to be down about 10% by year end (named-party figure). And it is trying to get paid for the read directly. Asked how it could extract value from 2.7 billion citations, Shah said monetising that relevance is the priority, that the company remains open to licensing deals, and that its litigation against OpenAI should bring clarity.
Notice what is missing from that list: any mechanism that pays Ziff Davis per AI impression. The metric exists. The market for it does not.
What does this mean for publishers?
First, measure the read. If you have Search Console access, the AI impressions figure is now the most important number in it, because it tells you how much of your content AI is consuming on your behalf. Put it next to clicks every month.
Second, do not mistake blocking for leverage. AI Overviews and AI Mode are fed by the same crawler as Google Search. Opting out of the read means opting out of the remaining referral too, and the content still circulates through every source that did not block.
Third, the industry-level numbers say this is not a Ziff Davis problem. The IAB Tech Lab puts publisher losses from AI bot traffic at around $2 billion a year (reported, IAB Tech Lab). OPA Europe and Rossel found one in fifty publisher visitors is now an AI bot (reported, OPA Europe / Rossel, January 2026).
What does this mean for advertisers?
The audience did not shrink by 37%. It moved inside the answer. Google's query volume is at a five-year high according to Shah's reading of Alphabet's own reporting, yet the click that used to deliver an ad impression on CNET now often ends at the AI summary.
For brands, that means the decision is increasingly made in a space where display and programmatic do not appear. Shah also argued that marketers are leaning towards trusted, well-cited brands as the open web fills with low-quality AI content. If the publisher is cited 2.7 billion times, the context it provides is reaching people even when the page is not.
The blankspace read Ziff Davis has just shown the market the size of the invisible audience: ten AI reads for every click, measured by Google. The referral was always a proxy for value. The read is the value. A publisher that waits for licensing deals and lawsuits to price 2.7 billion impressions is waiting on someone else's timetable. The faster route is to monetise the read where it happens, at the moment an AI agent requests the page, instead of hoping a shrinking share of those reads turns back into clicks.
Key figures
| Figure | Value | Provenance |
|---|---|---|
| Organic Google referrals to Ziff Davis, Q3 2026 YoY | down about 37% | Named-party figure (Vivek Shah, CEO) |
| Organic Google referrals, Q3 2025 YoY | down about 23% | Named-party figure (Ziff Davis) |
| Two-year decline in organic Google referrals | more than 50% | Named-party figure (Ziff Davis) |
| AI impressions in AI Overviews and AI Mode, Q3 2026 | roughly 2.7 billion | Named-party figure (Ziff Davis, citing Google Search Console) |
| AI impressions vs Google Search clicks | nearly 10x | Named-party figure (Ziff Davis) |
| Implied Google Search clicks, Q3 2026 | on the order of 270 million | blankspace estimate |
| Q3 2026 revenue (continuing operations) | $289.3m, down 5.6% | Reported (Ziff Davis 8-K) |
| Advertising and performance marketing revenue | down 7.7% | Named-party figure (Bret Richter, CFO) |
| Technology & Shopping revenue | $82.0m, down 3.8% | Reported (Ziff Davis 8-K) |
| Technology & Shopping goodwill impairment, Q3 2026 | $29.6m | Reported (Ziff Davis 8-K) |
| Share of revenue exposed to search traffic | about 13% | blankspace estimate (from company's one third / 40% framework) |
| Expected headcount reduction by year end 2026 | about 10% | Named-party figure (Ziff Davis) |
| Annual publisher losses from AI bot traffic | about $2bn | Reported (IAB Tech Lab) |
| Publisher visitors that are AI bots | 1 in 50 | Reported (OPA Europe / Rossel, Jan 2026) |
Primary source, opened and verified: Ziff Davis Q3 2026 earnings release (SEC 8-K, Exhibit 99.1) and Ziff Davis Q3 2026 earnings call transcript, 6 November 2026