It was a predicted threat. Now it's a measured fact: the massive deployment of Google's AI Overviews — these AI-generated summaries that appear at the top of the results page — reduces traffic to publisher sites by 34.5% on average when they are present.

This figure comes from a study by Ahrefs covering 300,000 Google searches. It's the moment when fear becomes fact. And for media, specialized blogs, niche sites, it's a paradigm shift in economics.

The Raw Data

-34.5% organic click-through rate when an AI Overview is present at the top of the page (Ahrefs, 300,000 queries)
Some independent studies measure drops up to -46% depending on the query categories
58% of Google searches now end without any external link being clicked — the zero-click searches
48% of queries now display an AI Overview (compared to 5% 18 months ago)
Small sites have lost -60% of search traffic in two years, compared to -22% for large publishers
Counter-trend: brand queries with AI Overview see their CTR increase by +18% (when someone searches “Le Monde covid” rather than a generic query)

What Has Mechanically Changed

Before AI Overviews, Google's mechanics were simple: the user types their question, Google displays 10 blue links, they click one or two. The publisher ranking in the top 3 would capture 40 to 60% of the traffic.

Since 2024-2025, the results page has been revamped:

A 3 to 6 paragraph AI summary directly answers the query
The cited sources are in small links next to or below the summary
The traditional 10 blue links are pushed to the bottom of the page — often below the fold

Result: in more than half of the cases, the user gets their answer without clicking. The traffic that used to pay publishers (advertising, subscriptions, indirect e-commerce) vanishes.

Ranking first on Google no longer guarantees a click. The new currency of visibility is being cited within the AI Overview.

Inequality Intensifies

The picture of losers and winners is stark:

Major media (New York Times, Le Monde, BBC) hold up: their brand is searched directly, they are cited in AI Overviews, and part of their traffic comes through proprietary apps and newsletters.
Niche sites and specialized blogs are bleeding: they thrived on long-tail SEO, precisely the segment where AI Overviews excel (“how to do X”, “what is Y”).
E-commerce sites hold steady — users still want to see and buy products — but editorial comparison sites and buying guides are devastated.

Counter Strategies Implemented

Publishers that survive — for now — combine several approaches.

1. Optimize to be Cited in the AI Overview

A new emerging discipline, often called AIO (AI Optimization) or GEO (Generative Engine Optimization). It's no longer about ranking #1, but about being chosen as a source by the model. This involves:

Writing direct and structured answers to common industry questions
Using precise, sourced factual data — LLMs prefer citing content that “looks” like verifiable facts
Structuring content with schema.org tags, tables, lists, clear definitions
Avoiding fluff and 300-word intros before the real answer — the AI extracts the substance and discards the rest

2. Bypass Google: Focus on Direct Relationships

Newsletters: Substack, Ghost, Beehiiv are booming. The email relationship is direct, measurable, monetizable, and out of Google's reach.
Mobile apps and push notifications
Podcasts and YouTube videos — formats less “cannibalizable” by a text AI summary
Closed communities (Discord, WhatsApp, Slack)

3. Sell Content Directly to AI Labs

Several major publishers (News Corp, Financial Times, Axel Springer, Le Monde, The Atlantic) have signed eight-figure deals with OpenAI, Google, Anthropic to license their content. It's a pragmatic conversion: since AI is going to scrape the content anyway, might as well get paid directly. However, these deals benefit mainly the big players — small publishers remain shut out.

4. Legal Action

The New York Times launched the first major lawsuit against OpenAI and Microsoft in 2023. Others are following. The class-action settlement against Anthropic ($1.5 billion in 2025 over pirated books) is a precedent being used for other battles.

5. Design for the Returning User

If the single click disappears, the loyalty relationship becomes central. Publishers investing in:

A recognizable editorial identity (voice, tone, visual style)
Known and followed author signatures
A consistent editorial schedule that builds a habit
Proprietary formats (long features, flagship podcasts, events)

...see direct traffic (accessing the URL without going through Google) soar, partially offsetting the SEO decline.

6. Diversify Beyond Text

Many publishers are pivoting to formats that AI summaries don't yet capture well:

Video (YouTube, TikTok, Instagram Reels)
Podcast (Spotify, Apple Podcasts, RSS feed)
Interactive reports, infographics, data visualizations
Paid physical events

7. Block AI Crawlers

Some publishers (Reuters, News Corp, several European media) are technically blocking training bots (GPTBot, Google-Extended, ClaudeBot). A radical choice: lose AI visibility in the medium term, but keep content out of competing models. The debate is unresolved — many analysts think it's a lose-lose situation.

The Landscape in 24 Months

Three scenarios are emerging:

Hard Consolidation. The 20 largest global publishers survive through licensing deals and their brand. Thousands of blogs and specialized sites gradually die off. The textual web ecosystem contracts violently.
Forced Regulation. The EU or other jurisdictions impose mandatory compensation for publishers by AI engines (extension of the Digital Services Act or a new specific directive). The market is forcibly redistributed.
Web Fragmentation. A significant portion of content leaves the open web for closed silos (apps, paid newsletters, communities) that are not indexable. Google becomes less relevant, a new engine emerges for this premium content.

The most likely: a mix of the three, with strong regional differences. What is certain, however: the model “I publish on my site, Google sends me free traffic, I monetize with ads” — the equation that supported the editorial web for 20 years — will not return.

For publishers, the time is for the rapid rebuilding of a model where the reader is a direct contact, not a stream of clicks bought from Google. The transition will be painful — it has already begun.

Tag: édition

Google Search: -34% Traffic to Publishers Due to AI Answers — Counter Strategies