Google Isn’t the Front Door Anymore. Your Blog Still Has a Job.

Two years ago Gartner predicted a 25% drop in traditional search volume by 2026, and every content team I know quoted it at least once in a planning deck, mine included. The year arrived, and Graphite, working with Similarweb across the top 40K US sites, found organic search traffic down 2.5% year on year, with the ten largest sites up about 1.6% and the losses bunched among sites ranked roughly 100 to 10K.

So the apocalypse didn’t happen, and anyone whose blog got flattened this year knows that’s a useless conclusion, because the aggregate was never the number that applied to them.

Where the organic traffic decline is really concentrated

Look at publishers instead of the whole web and the picture inverts: Chartbeat data covering more than 2500 news sites, published in the Reuters Institute’s 2026 trends report, showed Google organic traffic down 33% globally between November 2024 and November 2025, and down 38% in the US, and the same survey of 280 media leaders has them planning for search referrals to fall another 43% over three years, with one in five expecting losses above 75%.

These datasets are measuring different populations, and the variable separating them isn’t industry or site size so much as query type. Traffic that came from questions a paragraph can close has gone, while the queries where a searcher still needs the actual page held up more or less fine.

So the organic traffic decline reads as either a rounding error or a redundancy depending on who’s describing it, and “is organic traffic down” has no useful answer at that altitude. Ask instead which queries you lost, and whether you were ever monetising them.

I ran into my own version of this while auditing my Search Console data, and wrote it up in Impressions But No Clicks: 828 impressions, 14 clicks, on a query where I was visible and simply wasn’t needed. Nothing was broken; the page had already been read, in summary form, before anyone got as far as deciding whether to visit it.

The 67-word test

The Pew Research Center tracked the browsing behaviour of 900 US adults across roughly 69K Google searches in March 2025, which remains the cleanest behavioural dataset available on what AI summaries do to clicks. Users clicked a traditional result 8% of the time when a summary appeared, against 15% when it didn’t, and clicked a link inside the summary itself 1% of the time. They ended the session entirely on 26% of pages carrying a summary, against 16% without.

The number I keep coming back to from that study is smaller and more useful: the median AI summary was 67 words long.

Take the page that used to be your best performer, and try to write a complete answer to its headline in 67 words. If you can, Google can too, and it already has.

It takes about four minutes per page and it’s more diagnostic than any audit score, because it goes straight at what determines whether the page can be replaced by a paragraph.

A page passes when compression costs the reader something they came for: a dataset nobody else has, a decision that cost somebody real money, a comparison that ends in an actual recommendation, or a method detailed enough that following the 67-word version would get someone fired.

Which content types survive the organic traffic decline

Funnel stage is the wrong axis here, because plenty of decision-stage pages got eaten and plenty of top-of-funnel material held up fine. What separates them is compressibility.

Content type	What an AI answer does to it	Verdict	What to do
Definitional explainers (“what is X”)	Absorbed completely; the summary is the article.	Dead as a traffic play.	Keep one, as a definition source for citation, and stop producing more.
Generic how-to guides	Steps get lifted and reordered; no reason to click.	Mostly dead.	Rewrite around a specific implementation with the numbers and the mistakes.
Original research and proprietary data	Cited heavily, and the citation names you.	Strongest surviving format.	Produce more of it; this is where the budget should move.
Opinion with a real stake	Summarisable, not replaceable, because people want the person who holds it.	Resilient.	Take positions someone can argue with.
Comparisons that end in a recommendation	Pulled into shortlist answers constantly.	Resilient and commercially loaded.	Include the losing case or you won’t be trusted with the winning one.
Product, pricing, security and integration docs	Barely touched; the buyer still has to come and read the real thing.	Untouched.	Treat as pipeline infrastructure, not content.

The two rows I’d fight for are original research and the comparison. I’ve argued the case for the first in more detail in Original Research as Content Strategy, and the second is the reason the technical pages I built for CIO and CTO buyers at a B2B EdTech SaaS company kept converting while the awareness-stage library got quieter; I wrote about that build in Technical Content for CIOs.

Your blog now supplies the shortlist it used to compete for

The ROI conversation changes here, which is why I’ve stopped treating declining sessions as automatically bad news.

G2 surveyed 1076 B2B software buyers in March 2026 and found 51% now start their research with an AI chatbot. More striking: 69% chose a different vendor than they’d originally planned to based on what the chatbot told them, and a third bought from a vendor they’d never heard of before. Semrush’s survey of 519 B2B professionals using AI at work found 72% using it during early research and 61% comparing vendors directly inside the tool.

Read those two together and the blog’s job description changes from acquiring the visitor to supplying the answer that decides whether a visitor ever becomes a candidate at all, and a third of the time that’s happening for a company the buyer couldn’t have named an hour earlier. Which means an article that gets cited in a comparison answer and produces zero sessions did more for pipeline than a listicle that pulled 400 visitors with no buying intent behind them.

Getting cited is a separate discipline with its own rules, and I’ve documented mine in the AEO/LLMO Content Optimisation Framework. The TLDR is that substance earns the citation and structure only delivers it, so a green audit score on a thin page is a tidy way to stay invisible.

What to report when sessions fall and nothing is wrong

If your monthly report leads with traffic, the organic traffic decline will read as failure every month until someone kills the budget. Change what leads, and say why you’re changing it, which is that the measurement stopped describing the outcome some time in 2025.

• Citation share first. Run a fixed prompt set of the questions your personas actually ask, monthly, across ChatGPT, Claude, Perplexity and Gemini, and log whether you appear, who does instead, and what the answer says about your category. I’ve written up how I run this in AI Citation Tracking.
• AI referral quality goes next. The volume is still tiny, and it behaves differently from standard organic traffic. Segmenting AI referrals properly in GA4 takes about twenty minutes and it’s the fastest way to stop arguing about whether the channel matters; the setup is in Track AI Traffic in GA4.
• Branded search and self-reported attribution close it out. A rising branded query volume against falling non-branded traffic is the signature of a brand being discovered somewhere your analytics can’t see. Adding a free-text “how did you hear about us” field to the form is easy and consistently tells a different story from the CRM.

Traffic stays in the report as a diagnostic line. At the B2B EdTech SaaS company where I tracked 17+ KPIs a quarter, the two numbers that got quoted back to me in leadership meetings were marketing-sourced pipeline at 48% in peak months and €230K+ influenced in a single reporting period, and neither of them needed a sessions chart to make sense. I’ve laid out the full hierarchy in Content Marketing Metrics That Matter.

The limits

I’m reasoning from about eighteen months of post-AI-Overviews data and a handful of libraries, which isn’t enough to be certain about anything.

Citation tracking is genuinely bad as a measurement discipline right now. Results move between platforms and between runs of the same prompt, and there’s no shared standard for any of it, with tooling most of which didn’t exist two years ago.

I’d also push back on my own table. “Untouched” is doing optimistic work in that last row, and agentic buying assistants that read pricing and security documentation on the buyer’s behalf will change it. My bet is that those pages get more valuable while the session data around them gets stranger, and I’ve thought through what that does to content structure in Content for AI Agents.

What I’d do differently, looking back: I spent 2024 optimising for the click and should have started measuring citations six months before I did.

A blog that was a traffic channel and nothing else has lost a channel to the organic traffic decline, and you should say it so to whoever funds it. Where the blog was an argument with receipts attached, it’s now the raw material for every answer engine your buyers ask, and deciding which one you were running is worth an afternoon with a spreadsheet and the 67-word test, which is roughly where I’d start if I walked into your library cold. The follow-on question, whether the blog is really a blog at all any more, I’ve argued out in Stop Running a Blog, Start Running a Media Property.