The Annual Report · April 2026

The State of AI SEO 2026

The web is being rewritten. Here's who's winning and who's losing in the year search broke.

AI Overviews now appear on 25% of Google searches. ChatGPT has crossed 800 million weekly active users. Publisher traffic is down a third globally. And yet — ranking #1 in Google still gives you a 58% chance of being cited inside a ChatGPT answer. Seven chapters on how we got here, and what still works.

Start reading Jump to the full data room

~12 min read · 7 chapters · all data source-linked

Contents

01The Zero-Click Era Accelerates

02ChatGPT as a Search Engine

03The Citation Economy

04What Still Works: Front-loaded Content

05Schema, E-E-A-T & the New Ranking Factors

06The Publisher Collapse

07The Playbook for 2026

CHAPTER 01The Zero-Click Era Accelerates

The Search Page Is No Longer the Destination

AI Overviews broke the unwritten contract between Google and the open web — and neither side is going back.

In March 2025, AI Overviews appeared on roughly 13% of Google searches. By early 2026, that number had nearly doubled to around 25%. For information-seeking queries — the bread and butter of organic SEO for two decades — the SERP is no longer a list of ten links. It is a generated answer, with two or three citations tucked beneath it.

The click-through math got ugly fast. The average CTR on position 1 — the holy grail of SEO — fell from around 15% on a standard SERP to about 8% when an AI Overview appears above it. On informational queries, roughly 47% of what used to be clicks now stay inside Google itself.

Publisher traffic from Google fell by roughly one third across 2025. For informational content — the largest category on the open web — the drop was closer to half.

— Press Gazette, 2025 Annual Publisher Report

25%

Google searches with an AIO

15%→8%

Position 1 CTR with AIO

73%

users who don't click past AIO

−47%

Clicks on info queries

Seventy-three percent of users now report that an AIO answer is “good enough” that they don't feel the need to click through. That statistic, more than any other, explains why 2025 was the year the zero-click era stopped being a threat and became the baseline. This is the terrain every SEO strategy — every content plan, every traffic forecast, every backlink-building spreadsheet — now has to live on.

CHAPTER 02ChatGPT as a Search Engine

A Second Google Grew Up in 24 Months

OpenAI's weekly active user base doubled from 400M to 800M+ in under a year. And a growing share of those users are searching, not chatting.

In February 2025, OpenAI reported 400 million weekly active users. By late 2025, that figure had crossed 800 million, and the 1 billion WAU milestone was visibly in reach for early 2026. That is the fastest consumer-product growth curve ever recorded — faster than TikTok, faster than Instagram, faster than the original Google.

More importantly: the use case has shifted. When ChatGPT launched it was largely a chat product. Today, triangulating across Semrush, Similarweb, and SparkToro research, ChatGPT fields the equivalent of roughly 12% of Google's total search query volume. Somewhere between a third and half of ChatGPT prompts now look indistinguishable from classic Google queries — informational, navigational, transactional.

800M–1B

ChatGPT WAU

2×

User base growth in 12 months

12%

of Google's search volume

37%

of prompts are search-like

Perplexity, Claude, and Gemini add a second, third, and fourth pool of search demand on top of that — each with its own citation biases and surface patterns. The search market did not shrink in 2025. It multiplied. It just multiplied into interfaces that most SEOs weren't instrumented for.

Gartner's projection — that 25% of organic search traffic will shift to AI chatbots by the end of 2026 — looked aggressive when it was published. Two years in, it looks conservative.

CHAPTER 03The Citation Economy

The New Currency Is the Link Inside the Answer

Rankings haven't stopped mattering. They've just become the upstream signal for a new, higher-value outcome: the citation.

In 2025, seoClarity published the single most important data point of the year: 43.2% of pages ranked #1 in Google were cited by ChatGPT for the same query. Not 10%. Not 25%. Better than four in ten.

That headline was accompanied by an even more telling power curve. Position 1 had a 58% chance of being cited. Position 10 had a 14% chance. The citation opportunity isn't evenly distributed across the first page — it is brutally concentrated at the top.

If you rank #1 in Google, you have a 58% chance of being cited in ChatGPT. If you rank #10, you have a 14% chance. The distance between positions has never been worth more.

— seoClarity, LLM Citation Study 2025

43.2%

of #1 results cited in ChatGPT

58%

citation rate at position 1

14%

citation rate at position 10

5.2

avg citations per ChatGPT answer

Citations are the new backlinks. They are what drives the incremental trickle of referred traffic (which is now growing ~1% month over month across most sites). They are what drives brand discovery inside AI interfaces where the user never sees a SERP. And they are measurable — with the right tooling — across all four major LLMs.

The practical implication is the one most SEO teams still haven't internalized: the unit of work shifted from the page to the passage. LLMs don't cite pages. They cite paragraphs. 89% of LLM-cited passages are under 120 words. Your job is no longer to write a page that deserves a ranking. It is to write a page composed of ten tight, self-contained, citation-ready passages.

CHAPTER 04What Still Works

Front-Loaded Content Eats the Answer Box

44.2% of LLM citations come from the first 30% of a page's text. Everything important must happen above the fold.

A Semrush citation study published in 2025 surfaced the single most actionable finding of the AI SEO era: 44.2% of LLM citations come from the first 30% of a page's text. The traditional SEO instinct — bury the answer, build up context, reward the reader for their scroll — is now actively disqualifying your content from being quoted.

The pattern is consistent across ChatGPT, Perplexity, Claude, and Gemini. Large language models don't read to the bottom. They sample the top, extract the cleanest attributable passage, and move on. The writer who wins 2026 is the writer who leads with the answer, then earns the reader's scroll with depth, examples, and evidence.

44.2% of LLM citations come from the first 30% of the page. The top of the page is worth more than it has ever been — not for Google, but for the models that read on Google's behalf.

— Semrush, LLM Citation Study 2025

This is the single most teachable mechanical change an SEO team can make in 2026. Every article gets a 60–120 word summary paragraph immediately under the H1. Every H2 begins with a direct one-sentence answer before expanding into the analysis. Every list is self-contained, numbered, and quotable.

The upside is that the change also helps Google's classic featured snippet selection, PAA inclusion, and on-page time-to-answer metrics. There is no tradeoff. Front-loading is simply the correct structural move for the post-2024 web.

CHAPTER 05Schema, E-E-A-T & Ranking Factors

The Tags Got More Important, Not Less

FAQ schema triples your LLM citation rate. Author bylines double it. In an AI-first SERP, structured trust signals stopped being table stakes and started being leverage.

A running theme across 2025 citation research: LLMs disproportionately favor content that comes with structured trust signals. Pages with FAQ schema are cited roughly 3.1× more often than pages without. Pages with a visible author byline and accompanying bio are cited 2.4× more often than anonymously authored content. And 71% of all LLM-cited domains carry domain authority of 50 or higher.

None of this is surprising in isolation. What's new is the compounding. An AI-first SERP evaluates every candidate page against multiple trust axes simultaneously — schema, author credibility, domain history, recency, passage structure. A page that hits four of five boxes absolutely dominates a page that hits one.

3.1×

citation lift from FAQ schema

2.4×

citation lift from author byline

71%

of cited domains are DA 50+

The practical implications: every content page gets FAQPage schema on the FAQ section. Every article gets a Person-schema-linked author bio. Every fact-heavy claim gets a <cite> tag and a source. The work is unglamorous. It is also measurably rewarded by every major LLM's citation logic.

CHAPTER 06The Publisher Collapse

A Third of the Open Web's Traffic, Gone

Press Gazette's numbers are the blood on the floor. Publishers lost ~33% of their Google traffic in 2025. The ad revenue loss across the industry is estimated at $2.1 billion.

Press Gazette's 2025 cross-publisher analysis is the single most important business document of the AI SEO era. Across a broad sample of global news publishers — from the Guardian to Condé Nast to mid-market regional outlets — Google-sourced traffic fell by roughly one third during 2025. For definition-style and “what is” content, the drop was closer to 58%.

The cascading effects reached the ad stack within six months. Display RPMs on informational content fell ~6% across Mediavine and Raptive networks. Analyst estimates place the total annual ad revenue loss across the global publishing industry at around $2.1 billion USD. For newsrooms already under pressure from platform shifts, this was the second existential shock in a decade.

Global publishers lost approximately one third of their Google organic traffic in 2025. The ad revenue impact has been estimated at roughly $2.1 billion.

— Press Gazette, 2025 Annual Publisher Report

Twenty-two percent of publishers now block GPTBot or OAI-SearchBot via robots.txt. The political logic is obvious. The SEO logic is more complicated: blocking the crawler also blocks citation eligibility, and citations are — increasingly — the only AI-referral mechanism left.

The honest read on 2025: the publisher business model broke on Google's side before it had a chance to be rebuilt on the AI side. 2026 is the year that rebuild either happens or doesn't.

CHAPTER 07The Playbook for 2026

10 Tactics That Actually Work

Not theory. Not trend-pieces. Ten concrete, defensible, measurable changes to make to your SEO program this quarter.

Front-load the answer

44.2% of LLM citations come from the first 30% of page text. Lead with the direct answer, then earn the deep-dive with depth, examples, and data.

Write for the passage, not the page

LLMs quote paragraphs, not articles. Every 80–120 word block should be a self-contained, attributable answer — tightly scoped and quotable.

Ship FAQ schema on everything

Pages with FAQ schema are cited by LLMs roughly 3× more often. It's the cheapest, highest-leverage GEO tactic available today.

Make the author real

Author bylines double citation probability. Add a visible byline, link to a real bio page, reference prior work, and implement Person schema.

Publish fast, update faster

50% of Perplexity citations are 2025 content. Keep a recency flag — updated dates, revision history, fresh stats — on your highest-priority pages.

Treat Reddit like a channel, not an enemy

18% of ChatGPT citations come from Reddit and forums. Own threads in your niche. Build subreddit presence where your buyers research.

Optimize for the long tail

AIOs trigger 3× more often on 7+ word queries. Long-tail informational content is both the biggest AIO risk and the biggest citation opportunity.

Build internal linking that tells a story

LLMs follow link graphs to assemble context. Every new article should pull its weight in the link graph — both incoming and outgoing.

Measure citations, not just rankings

Rankings remain the leading indicator. Citations across ChatGPT, Perplexity, Claude, and Gemini are the lagging one. Track both.

Don't abandon Google

43.2% of #1 Google results are cited in ChatGPT. Organic ranking is still the single strongest predictor of LLM citation. SEO didn't die — it got upstream.

FAQ

Frequently asked questions

What is the state of AI SEO in 2026?

AI search is now roughly 25% of all Google queries (via AI Overviews), and ChatGPT handles the equivalent of 12% of Google's total search volume with 800M–1B weekly active users. Publisher Google traffic fell by roughly one third in 2025. SEO hasn't died — it's fragmented into Google SEO, LLM citation optimization (GEO), and AI Overview optimization.

Is SEO still worth doing in 2026?

Yes — and more so than ever, because SEO is now upstream of LLM citations. 43.2% of pages ranked #1 in Google are cited by ChatGPT; position 1 has a 58% citation rate versus 14% for position 10. If you rank in Google, you are significantly more likely to be cited by AI. The classic playbook hasn't died; it's been extended.

What's the single biggest change from 2025 to 2026?

The collapse of the zero-click assumption. In 2025, the question was 'will AIOs reduce my clicks?' In 2026, the question is 'how do I get cited inside the AIO or LLM response itself?' The entire optimization target has moved from the SERP to the generated answer.

Which content types are most at risk in 2026?

Thin definitional content, 'what is X' pages, and short how-tos are losing the most traffic — down 47–58% on informational queries. Deep editorial content, original research, expert analysis, and tool-backed comparisons are gaining relative share because they are harder to summarize.

How do I optimize for ChatGPT and Perplexity citations?

Front-load your answers (44.2% of citations are pulled from the first 30% of text), ship FAQ schema, add visible author bylines and Person schema, publish recent content (especially for Perplexity which heavily favors 2025 content), and keep ranking well in Google — which remains the strongest single predictor of LLM citation.

What does Gartner predict for 2026?

Gartner projects that by the end of 2026, 25% of what was previously organic search traffic will shift to AI chatbots — ChatGPT, Perplexity, Claude, and Gemini. That's not a future forecast; industry data through early 2026 is on-pace to meet or exceed it.

Is Harbor built for this new reality?

Yes. Harbor writes 3,500+ word, front-loaded, FAQ-schema'd, author-bylined, internally-linked articles designed specifically to rank in Google and be cited by LLMs. It's the SEO writing tool built for 2026, not 2022.

Built for this new reality

Harbor is built for 2026.
Not 2022.

Front-loaded articles. FAQ schema baked in. Author bylines. Internal linking. Fact-checked against live sources. Designed to rank in Google and be cited by ChatGPT — because in 2026 those are the same job.

Start free See the full data room