Three publishers filed their numbers. Only one rebuilt the revenue
Everybody quotes the traffic decline. Almost nobody reads the filings underneath it, which is a pity, because the filings say something far more useful about where the money actually went.
What three filings say that the coverage did not#
The fox does not hunt where the noise is. It hunts where the tracks are, and in this story the tracks are filed with the Securities and Exchange Commission rather than published in a trend piece. Any content marketing strategy built this quarter should start there.
Three publishers reported Q2 2026 within two days of each other, and their results diverge so sharply that the divergence is the whole content marketing strategy lesson. Take them in order.
Why does a publisher's balance sheet belong in a content marketing strategy discussion at all? Because publishers are the only operators who report, in public and under audit, what happens to a business when search referral falls away. Every brand running a blog is running a small version of the same experiment without the disclosure requirements.
The New York Times Company grew. Its Q2 2026 results filing reports total revenues of $762.5 million, up 11.2% year on year, with digital advertising revenues of $114.0 million, up 20.7%, attributed to "strong marketer demand and growth in advertising supply".
Ziff Davis, a portfolio of ad-funded properties, went the other way. Its Q2 2026 filing shows revenues down to $286.7 million from $294.8 million, and an operating loss of $44.7 million against operating income of $13.8 million a year earlier.
| Publisher | Q2 2026 revenue | Direction | Where the money comes from |
|---|---|---|---|
| New York Times Co. | $762.5m | Up 11.2% | 13.35m subscribers, digital ad up 20.7% |
| USA TODAY Co. | $536.3m | Digital now 47.4% of total | Digital-only ARPU up 34% to $10.47 |
| Ziff Davis | $286.7m | Down 2.7%, operating loss $44.7m | Ad-funded portfolio across four segments |
The segment detail is where it stings#
Ziff Davis reports four segments and three of them are flat or falling. Technology and Shopping fell 5.0% to $76.7 million. Health and Wellness fell 4.8% to $94.7 million. Gaming and Entertainment rose 0.9% and Cybersecurity and Martech rose 0.5%, which is stability rather than growth.
Those are exactly the categories that were built to capture commercial search intent. When the intent stops arriving as a click, a portfolio of review and comparison properties has no second door for the reader to come through.
The number everybody quoted, and where it is not#
One figure travelled everywhere this week: that around 50% of relevant queries now trigger an AI Overview, up from roughly 36% a quarter earlier, attributed to Ziff Davis. It is a striking number and it may well be accurate. It is not in the filing.
The phrase "AI Overview" appears zero times in the Ziff Davis Q2 2026 results release. There is also no "advertising and performance marketing" revenue line in that document. Both figures come from the earnings call and the accompanying slides, not from the filed press release the coverage cites.
There is a better source for the underlying question anyway, and it has a published method. The Pew Research Center ran a controlled study of 900 US adults who agreed to install a browsing tracker, capturing 68,879 unique Google searches, of which 12,593 produced an AI summary.
Read those four bars together, because the pair at the end matters more than the pair at the start. Clicks roughly halve, which everybody reports. Sessions ending rises from 16% to 26%, which almost nobody reports, and that is the number that should reshape a content marketing strategy.
A reader who ends the session did not go to a competitor. They finished. The Pew study also notes users clicked a link inside the summary on just 1% of visits, and states its own limitation plainly: it covers Google only.
What actually replaced the referral#
The publishers who grew did not find a clever new traffic source. They raised the value of the readers they already had, which is a slower and much duller answer than the one most conference talks offer.
The New York Times added roughly 280,000 net digital-only subscribers in the quarter to reach 13.35 million, with digital-only ARPU up 3.1% to $9.94, and affiliate, licensing and other revenues up 7.1%. Growth came from more readers and more money per reader at the same time.
Digital-only subscribers
Up roughly 280,000 net in the quarter.
Digital-only ARPU
Up 3.1% year on year. More readers and more per reader.
Digital ad growth
Attributed to marketer demand and advertising supply.
USA TODAY Co. tells the same story in a harder market. Its Q2 2026 filing reports digital-only subscription revenues of $45.6 million and "the second consecutive quarter of year-over-year growth", with total digital revenues of $254.3 million, "or 47.4% of total revenues".
The ARPU move nobody talks about#
The most instructive figure in that filing is the average revenue per user. Digital-only ARPU for USA TODAY Co. rose 34% to $10.47, and for USA TODAY Media specifically it rose 39% to $11.03. Newsquest, its UK arm, fell 5% to $5.70.
A 34% ARPU rise while unique visitors sit at 158 million is a business deliberately trading reach for depth. That trade is available to almost every brand publishing content, and almost none of them make it, because reach is the number that looks good in a monthly report.
Reach is the metric you report. Depth is the metric that pays. The gap between them is where most content budgets quietly die.
The uncomfortable implication for any content marketing strategy is that the winning move was not a content move. Nobody in these filings out-published their way back to growth. They changed who the reader was to them, from an anonymous session to a named, paying, reachable person, and the content served that change rather than the other way round.
Trade coverage caught the direction of travel accurately. Digiday framed it as publishers replacing Google with a little bit of everything, and PPC Land put the production cost side at roughly $2 billion a year across about 500,000 works.
Porting this into a content marketing strategy#
A brand is not a newspaper and does not sell subscriptions. The transferable part is not the business model, it is the diagnosis: a publishing operation whose only distribution is search referral has one point of failure, and that point failed.
Most content marketing strategy examples circulating in decks still assume the funnel starts at a search result. That assumption was reasonable for fifteen years. The Pew numbers and three filings say it now needs an explicit second and third route, owned and measurable.
Rebuild the content marketing strategy around routes rather than formats and the planning conversation changes shape immediately. The question stops being which posts to write this quarter and becomes which of the five routes below currently has no owner, no budget and no number attached to it.
Email, app or account. The only channel where the audience list is yours and no intermediary can reprice access to it overnight.
Raise value per reader rather than chasing volume. USA TODAY Co. lifted digital ARPU 34% while reach stayed flat.
Content that can be licensed is content with a revenue line that does not depend on anyone clicking it at all.
Publish where the audience already is rather than only linking back, accepting the traffic loss in exchange for reach that survives a ranking change.
Original data, research and named methods. It is the material that gets cited, quoted and linked when summaries replace listings.
That last route is the one a content marketing agency can build fastest for a client, because most businesses are sitting on proprietary data they have never published. It is also the route with the longest half-life, since a cited figure keeps working long after the page that carried it stops ranking.
The pressure is not evenly spread. PPC Land reports small publishers down 60% in search referrals over two years against 22% for large ones, and separately that 300 French dailies have filed a competition complaint seeking licensing remuneration rather than removal of the feature.
Five moves worth making this quarter#
A content marketing strategy that survives this does not require abandoning search. It requires refusing to let search be the only way anybody reaches the work, and building the measurement that proves the second route exists.
Treat this as a content marketing strategy audit rather than a campaign brief. Four of the five moves cost editorial time rather than media budget, and the one that costs money, native distribution, can be tested at a fraction of what a search-dependent programme already spends.
A workable content marketing strategy framework here has five parts, and they map onto the routes above rather than onto channels. Own the relationship, deepen the value, publish original evidence, distribute natively, and measure citation alongside clicks.
Start with the one that costs nothing#
Publishing original first-party data is the cheapest of the five and the most neglected. Most businesses already hold proprietary numbers, and the work is editorial rather than technical: choose a question, state the method, publish the figure with its sample size.
That is precisely what makes a page quotable when a summary sits above the results, and it is the discipline behind every piece folkfox ships, from referral collapse to the health media write-down. The scent a fox follows is a specific one.
Adjectives, no anchor
Publishers are seeing significant declines in search traffic as AI summaries reshape how audiences discover content across the web.
Figure, method, source
Pew Research tracked 68,879 Google searches across 900 US adults and found clicks on traditional results fell from 15% to 8% when an AI summary appeared.
The second version is the one a model repeats and a journalist cites, because it survives being lifted out of its paragraph. That is the entire test now, and it costs one extra sentence of rigour.
There is a wider pattern worth watching in the undergrowth. YouTube doubled its Partner Programme watch-hour threshold for new channels from February 2027, and Substack began showing readers an AI-detection score beside posts. Platforms are raising the bar on who gets paid and who gets believed, in the same fortnight.
If you want the first-party research programme, the distribution map and the citation measurement built rather than described, that is what folkfox content marketing does, alongside the brand strategy work that decides what is worth publishing at all. Quiet, quick, and finished before the next quarter files.
Frequently asked questions#
What are the 5 pillars of content strategy?
In this reading they are: an owned direct relationship with the audience, depth of value per reader rather than raw reach, original first-party evidence, native distribution on the platforms where the audience already sits, and measurement that counts citation as well as clicks. The pillars are routes to the audience, not content formats.
Is search traffic really collapsing for everyone?
No, and the filings show it. The New York Times grew revenue 11.2% in Q2 2026 while Ziff Davis posted an operating loss. The exposure is concentrated in publishers whose only distribution was commercial search intent, and PPC Land reports small publishers down 60% in referrals against 22% for large ones.
How much do AI summaries actually reduce clicks?
Pew Research Center tracked 68,879 Google searches from 900 US adults and found users clicked a traditional result on 8% of visits where an AI summary appeared, against 15% where it did not. Users clicked a link inside the summary on only 1% of visits. The study covers Google only.
Where should a brand start if search referral is falling?
Start with original first-party data, because it is the cheapest of the five routes and most businesses already hold the numbers. Choose one question, state the method and sample, and publish the figure. It creates material that gets cited even when the page itself is not clicked.
Do I need a content marketing agency to do this?
Not necessarily. The first-party research move is editorial work that an in-house team can do if it has access to the data and permission to publish it. A content marketing agency helps most with the measurement layer and with distribution mapping, which are the parts that usually stall internally.
What content marketing strategy examples show this working?
The clearest public examples are in the filings themselves. USA TODAY Co. lifted digital-only ARPU 34% to $10.47 while reach stayed broadly flat, and the New York Times grew subscribers and revenue per subscriber simultaneously. Both traded volume for depth deliberately.
Read more on this topic#
The archive that stopped receiving questions
What referral collapse looks like from inside a content archive.
Read the pieceA $54.8m write-down said what the revenue line would not
The balance sheet version of the same story, one quarter earlier.
Read the pieceThree companies, one week, and the lease nobody read
Why renting your distribution is the risk underneath all of this.
Read the pieceYour robots.txt is a request, and the fetchers know it
Today's other publishing story, on who is reading the content you publish.
Read the piece
Want a second route to your audience built, not described?
folkfox builds first-party research programmes, distribution maps and citation measurement for brands whose content strategy has one point of failure. Evidence you own, published so it gets quoted.