Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Search Traffic Decline: Publishers Face a Post-Search Rebuild, Not a Blip

Search referral traffic to publishers is falling sharply, with one recent report measuring a 40% year-over-year drop, and the industry's response is now about rebuilding distribution rather than waiting for a recovery.

Media & internetExplainerGrace OkonkwoPublished: 29 September 20264 min readSources 5
Search Traffic Decline: Publishers Face a Post-Search Rebuild, Not a Blip

On 28 September, Digiday's recap of its September Publishing Summit framed the industry's task as rebuilding for a "post-search era." Two years ago that phrase would have sounded alarmist. Now it reads as a planning assumption.

The dossier's recent headlines point the same way. One report from late September measured Google Search traffic declines for publishers at 40% year over year. Earlier in the month, Microsoft said it was adding an "ask a follow-up" feature to Copilot results that could further affect publishers' traffic. In March, reports described small websites seeing drops of up to 60%, and a separate antitrust filing alleged Google cannibalizes publisher traffic.

Court Filing Quotes OpenAI Engineer: Users "Won't Click" Links

That last item, from a court filing reported by Search Engine Journal on 28 September, cuts to the heart of the shift. If users increasingly expect answers rather than links, the traffic model that sustained much of the open web for two decades is structurally weakened, not merely having a bad quarter. Cloudflare's Matthew Prince made the related argument the same day, telling The Next Web that "bots don't click on ads." That line captures the mismatch between AI crawling and advertising economics.

The data on how people actually use AI for news is still thin. Statista examined AI tool usage in news consumption on 28 September. MediaPost profiled Forbes' chief innovation officer on how that publisher is adapting its reader relationship. Neither offers a clean replacement number for the lost search referrals.

Where the traffic is going, and who is losing most

Press Gazette's ranking of the 50 biggest UK news websites, published on 28 September, showed the Oxford Mail leading audience growth and tripling time spent in August. That is a useful counterexample: some titles are still growing, which suggests the decline is not uniform and depends heavily on brand loyalty, local relevance and direct audience relationships.

Small publishers are not so lucky. Reports from March described smaller sites hit hardest by search traffic declines, and a February antitrust filing alleged Google cannibalizes publisher traffic. Hyperallergic's move to paywall its content, reported by Artforum on 28 September, is one response: convert the readers you have rather than chase the ones the algorithm no longer sends.

Google's own product direction keeps pushing the same way. AI Overviews expanded across search results in June, and in March the company was reported to be rewriting headlines with AI, a change with direct consequences for SEO and click-through. Google has reported that AI Mode surpassed one billion monthly users, according to a Manticore Search blog post, and that people are asking longer, more complex questions that previously did not fit conventional search.

That shift in query behaviour is the mechanism behind the traffic loss. A query that once produced ten blue links now produces a synthesised answer. The publisher whose reporting informed that answer may receive nothing.

What publishers are actually doing about it

The rebuild is taking several forms. Digiday's summit coverage described publishers diversifying into newsletters, events, direct subscriptions and commerce. Press Gazette's audience data suggests local and community titles with strong habitual readership can still grow. Others are experimenting with licensing content to AI companies, though the dossier does not contain confirmed deal terms for any publisher.

There is a measurement problem underneath all of this. As Manticore Search's Klim Todrik wrote, full-text search still performs well with exact names, SKUs, product codes, brands and keywords, but it no longer gets users to an answer, which is what AI-assisted search now promises. Publishers selling search visibility are therefore selling into a system whose output increasingly bypasses them.

In a related corner of the same problem, researchers at Loro described the gap between syncing an index and syncing the content behind it: a user sees a new-message notification, opens the conversation and finds no new message, because the index synced but the body did not. The publishing analogue is a reader who sees a headline in an AI summary, cannot reach the article, and never becomes a visitor.

Scepticism is warranted about any single figure. The 40% decline, the 60% drop for some small sites and the growth at the Oxford Mail cannot all describe the same population. They describe different baselines, different verticals and different measurement methods. No source in this dossier reconciles them.

What is clear is the direction of travel. The industry's own conferences, trade press and platform announcements all point to less referral traffic and more reliance on direct relationships. The publishers still growing are those that already had one.

Comments 0

Sources

5
  1. 01Full-Text Search Still Works. It Just Doesn't Get You to an AnswerEN
  2. 02CRDTs are not enough: From CRDTs to a local-first sync engineEN
  3. 03The Macroeconomic Effect of AI: Sizing the Software Engineering ChannelEN
  4. 04The Intelligent Data Platform for the AI EraEN
  5. 05Routing LLM traffic across inference providers by cost, speed and reliabilityEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.