# SEO trends, autumn 2026: what the world's leading analysts are actually saying

> 68% of Google searches now end without a click, Google says optimizing for AI 'is still SEO', and the people who measure this for a living disagree on almost everything else. Rand Fishkin, Lily Ray, Kevin Indig, Aleyda Solís, Mike King, Cyrus Shepard and Glenn Gabe — quoted verbatim, with sources and dates.

_Source: https://seomatrix.ai/blog/seo-trends-autumn-2026/ · Updated: 2026-09-24_

---

Every autumn the industry produces a hundred "SEO trends" posts, and most of them are one person's opinion dressed up as consensus. This one is different in a boring way: every claim below is a **verbatim quote or a dated figure from a primary source**, collected on 24 September 2026, and the people quoted are the ones the rest of the industry actually reads — Rand Fishkin, Lily Ray, Kevin Indig, Aleyda Solís, Mike King, Cyrus Shepard, Glenn Gabe, Barry Schwartz — plus Google's own words.

Three numbers frame the season. **68.01%** of US Google searches ended without a click in January–April 2026 (SparkToro × Similarweb, 8 June). AI Overviews cut the click-through rate of the first organic result by **58%** (Ahrefs, 4 February). And ChatGPT's share of generative-AI web traffic fell from **~76% to ~53%** in a year (Similarweb, 29 July). Everything the analysts say is a reaction to those three facts.

## 1. Zero-click is the norm — but the number depends on who is counting

Rand Fishkin's June study is the one everyone cites:

> In the first four months of 2026, a whopping 68.01% of Google searches ended without a click.
> — Rand Fishkin, SparkToro, [8 June 2026](https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/)

His conclusion is not "give up on SEO", and it is worth quoting in full because it is usually shortened into the opposite:

> Much as it pains me to say, there's not much point (nor any hope) of fighting back by simply getting better at SEO. Don't take that the wrong way — your SEO still matters as much or more than ever before, it just won't earn you traffic the way it once did.
> — Rand Fishkin, SparkToro, [8 June 2026](https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/)

The same study kills a popular theory: AI Mode is not what took the clicks. "Only 0.34% of searches made their way to AI Mode from January–April 2026." The clicks went to AI Overviews, which sit on 20%+ of results pages and cut CTR by nearly 60% when they appear.

The number itself is contested. Mike King's team ran the same question on a different panel — 13.1 billion events from Datos — and got **46.96%** zero-click for October 2024 – December 2025 ([iPullRank, 30 July](https://ipullrank.com/zero-click-behavior-analysis-q3-2026)). Different panel, different period, different definition. Kevin Indig, who predicted the worst, corrected himself in July:

> I predicted up to 70% of 2024 organic traffic could be gone by 2026, but that was overeager. In the end, the numbers land at ~33% referral loss year over year, with 68% of Google searches now ending without a click.
> — Kevin Indig, Growth Memo, [27 July 2026](https://www.growth-memo.com/p/ai-halftime-report-h1-2026)

> **How to read the two numbers** — 68% and 47% are both real. One is US Google, January–April 2026, Similarweb panel; the other is a Datos panel over fifteen months. Quote either with its period and panel, never as "the" zero-click rate. If your own Search Console shows impressions rising and clicks flat — the "alligator graph" Fishkin describes — you are living the trend regardless of which average is right.

## 2. "GEO is just SEO" — Google made it official, the practitioners added footnotes

In May Google published a guide on optimizing for generative AI features, and its position is unambiguous:

> From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.
> — Google Search Central, [AI optimization guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide), announced 15 May 2026

The same document tells you to "prioritize effective SEO strategies over 'AEO/GEO hacks'" and names the hacks: "chunking" content, "unnecessary AI text files (like llms.txt)", "inauthentic mentions". On llms.txt specifically: "Google Search ignores them."

Lily Ray, who has spent the summer reverse-engineering ChatGPT's fan-out queries, lands in the same place from the other direction:

> I think analyzing fan-out queries matters because ChatGPT is using search engines to retrieve the results it uses to formulate an answer, which means this is fundamentally an SEO problem.
> — Lily Ray, [17 August 2026](https://lilyraynyc.substack.com/p/what-we-can-learn-from-evolving-chatgpt)

And, in a September interview:

> If you understood how AI search works, you would understand that SEO is more important than ever. The single best and most sustainable and important thing you can do for AI search visibility is good SEO.
> — Lily Ray, Search Engine Land, [2 September 2026](https://searchengineland.com/lily-ray-search-engine-land-20-for-20-interview-486614)

Glenn Gabe adds the mechanism that makes this true — AI answers are grounded in Google's index, so a Google penalty propagates:

> Drop in Google and you can drop across AI Search (since AI search can be grounding its answers via Google's index)
> — Glenn Gabe, GSQi, [10 June 2026](https://www.gsqi.com/marketing-blog/core-roars-back-google-may-2026-core-update-analysis/)

The footnote comes from Cyrus Shepard, whose August piece is the most balanced treatment of the question:

> Does your business need a separate and dedicated GEO strategy to win AI visibility? For the vast majority of websites, almost certainly not.
> — Cyrus Shepard, Zyppy, [4 August 2026](https://signal.zyppy.com/p/seo-strategies-for-ai-search)

> That said, the reverse statement we sometimes hear, "GEO is just SEO," is harder to defend.
> — Cyrus Shepard, same post

His reason: new surfaces keep appearing that classic SEO never had to think about — personal context in Gemini, assistants that filter sources the way a spam filter would. The work is 90% SEO. The remaining 10% is new, and it is where the arguments live.

## 3. Brand and trust outweigh every AI-specific tactic — and most citations are not on your site

Kevin Indig's half-year report is the sharpest reframing of the season:

> AEO/GEO is a brand channel disguised as a performance channel. AI recommendations shape demand (not citations). The unit of optimization is whether AI names, trusts, and recommends your brand.
> — Kevin Indig, Growth Memo, [27 July 2026](https://www.growth-memo.com/p/ai-halftime-report-h1-2026)

> Trust is the highest currency in AI search.
> — Kevin Indig, same report

Shepard's September survey of 130+ experts says the same thing with numbers: "Brand and Trust outweigh AI-specific factors: Experts generally agreed that what Google knows about and trusts your brand weighed more heavily than AI-specific tactics like 'chunking' or llms.txt files." ([Zyppy, 16 September](https://signal.zyppy.com/p/google-ai-ranking-factors))

Aleyda Solís supplies the uncomfortable corollary. In her SaaS study, **84–93% of "citation weight"** in ChatGPT and AI Mode came from third-party sites, not the brand's own:

> AI search is a 3rd-party citation problem with an on-page corroboration base
> — Aleyda Solís, [20 September 2026](https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/)

> The pages you get clicks on are usually cited; the overwhelming majority of the pages doing the citation work never receive a visit.
> — Aleyda Solís, same checklist

> Publishing authoritative content on your own site is necessary but rarely sufficient.
> — Aleyda Solís, [16 July 2026](https://www.aleydasolis.com/en/ai-search/saas-ai-search-optimization/)

Amanda Natividad, Fishkin's co-author, compresses the strategy into one line: "Your goal is to become a named entity, not get a bunch of pages to rank." ([SparkToro, 10 June](https://sparktoro.com/blog/zero-click-search-what-still-works/))

## 4. The measurement crisis: impressions without clicks, precision without accuracy

On 3 June Google launched the Search Console report for generative AI surfaces; on 31 August it went global, together with a control to opt out of AI Overviews and AI Mode. Barry Schwartz's reaction to the launch:

> Oh, and no, the reporting does not show click data - not that any of you expected that.
> — Barry Schwartz, Search Engine Roundtable, [3 June 2026](https://www.seroundtable.com/google-ai-performance-report-blocking-controls-41443.html)

Glenn Gabe, after two months of trying to separate AI Mode queries from the rest: "AIO and AI Mode data is there, but it's not broken out by surface. It's almost like Google wants to confuse site owners. :)" ([GSQi, 13 August](https://www.gsqi.com/marketing-blog/how-to-find-ai-mode-queries-gsc/))

Third-party "AI visibility" tools do not fix this, and Mike King wrote the definitive explanation of why:

> In AI Search, "accuracy" is a category error. The only thing you can have is precision, and if you don't understand the difference, you are going to make expensive decisions based on noise.
> — Mike King, iPullRank, [17 September 2026](https://ipullrank.com/accuracy-vs-precision)

> AI search tracking is a polling problem. The industry keeps trying to sell it as a scale.
> — Mike King, same post

The study he cites (Mike Sonders) ran one prompt 100 times in ChatGPT and saw about 44 different brands; only about 5 appeared in 80% or more of runs. A single check is not a measurement — it is one poll respondent. Fishkin's answer to the same problem is to stop chasing traffic as the metric at all:

> When we talk about that alligator graph, right, of impressions going up and traffic going down, clicks declining, guess what? Those impressions still count.
> — Rand Fishkin, SparkToro, [14 August 2026](https://sparktoro.com/blog/does-your-website-still-matter-in-the-zero-click-era/)

> **What this means for your dashboard** — Any tool that shows you "your AI visibility is 37%" from a single run per prompt is showing noise. Track a fixed prompt set, repeat runs, and report ranges and stability, not a single number. Our own tracker moved to repeated probes for exactly this reason — the first time we measured, a page we ranked #1–3 for on 68 queries with an AI Overview was cited in zero of them.

## 5. The assistant market is fragmenting — and llms.txt is still unread

Similarweb's July numbers reshape the "optimize for ChatGPT" plan: ChatGPT's share of generative-AI web traffic fell from **~76% to ~53%** between June 2025 and May 2026, Gemini grew to **~27–28%**, Claude to **~9%** ([Similarweb, 29 July](https://aisearch.similarweb.com/blog/gen-ai-stats/)). Indig sees the same shape on a different methodology (78% → 56%). Meanwhile ChatGPT answers *with citations* grew from 1.6% to 6.8% of prompts, and after the 7 May update the share of ChatGPT referrals landing on home pages jumped from 26–29% to 62–63% — the assistant increasingly sends people to the brand, not the page.

On the machine side, Cloudflare Radar's data reported in June put bots at **57.5% of HTML requests** — for the first time above human traffic. And the file everyone built for those bots is still unread: an Ahrefs study found that **97% of llms.txt files received zero requests** in May 2026 ([Ahrefs, 15 June](https://ahrefs.com/blog/llmstxt-study/)). Google's John Mueller, quoted by Search Engine Journal:

> I don't think anyone knows – it's purely speculative for now (the file has existed for years, yet none of the AI systems use it — what does it mean?).
> — John Mueller, via Search Engine Journal, [2 June 2026](https://www.searchenginejournal.com/google-says-llms-txt-is-purely-speculative-for-now/577576/)

His practical advice in the same thread is the one line that matters for agents: "the most basic agentic optimization… namely: don't block agents. I think that hurdle will be the biggest, for most sites."

## 6. Commodity content is dying, and the GEO-spam reckoning is coming

Google's guide draws a line between "commodity" and "non-commodity" content. Shepard's April analysis of 400+ sites that won the new Google puts it bluntly:

> In short, Google has moved beyond simply ranking "good content" to proactively rewarding what AI can't replicate.
> — Cyrus Shepard, Zyppy, [9 April 2026](https://signal.zyppy.com/p/winning-google)

The winners share a profile: a product or service, a task the visitor can complete on the site, proprietary assets, a narrow topic, a brand. Indig's August study of listicles found "no universal demotion" but a lot of churn — "only 52.5% of queries kept the same position-one URL between January and August", and "the clearest signal was freshness" ([Growth Memo, 31 August](https://www.growth-memo.com/p/what-makes-a-good-listicle)).

The flip side is risk. Gabe:

> There is a huge risk for site owners using risky tactics to game AI Search since if you get hammered by Google for doing that, your site will drop out of the 10-blue links, AI Overviews, AI Mode, and Gemini.
> — Glenn Gabe, GSQi, [10 June 2026](https://www.gsqi.com/marketing-blog/core-roars-back-google-may-2026-core-update-analysis/)

Lily Ray goes further, imagining "the GEO atomic bomb update" that removes experimenters from Google Search entirely — "it sounds extreme, but there's precedent for this." Solís lists the tactics that will get caught: "Prompt injection, fabricated reviews, manufactured community signals and citation gaming might end up producing short-term improvements, but they will get eventually addressed, and will trade durable trust for visibility that can backfire."

For the record, Google's 2026 update calendar per the Search Status Dashboard: a Discover update (February), spam updates on 24 March, 24 June, 18 August and 24 September, and two core updates — 27 March and 21 May. **There was no August 2026 core update**, whatever some trend posts say.

## Where the analysts disagree

| Question | One side | The other |
|---|---|---|
| Is SEO dead? | Fishkin: SEO "just won't earn you traffic the way it once did" | Ray: dropping SEO for AI search is "completely blatantly false"; SEO "matters more than ever" |
| Is GEO just SEO? | Google: "still SEO"; ignore llms.txt and chunking | Shepard: "GEO is just SEO" is "harder to defend"; Indig: it is a brand channel, not performance |
| How big is zero-click? | SparkToro/Similarweb: 68% (Jan–Apr 2026) | iPullRank/Datos: 47% (Oct 2024–Dec 2025) |
| Block AI bots? | 33% of SEOs in SER's poll say they would | Schwartz doubts it; Mueller: "don't block agents" |
| Is llms.txt worth it? | Google ignores it; 97% of files get no requests | It costs nothing; agent-readiness (markdown versions, accessibility tree) is rising |
| Is AI traffic valuable? | Adobe (via Natividad): AI referrals convert ~42% better | Ahrefs: Wil Reynolds' ChatGPT citations grew 1,900% with no visible business effect |

## What to do this quarter

1. **Measure the alligator, not the average.** Put impressions and clicks on the same chart per page; report the gap. Use Search Console's AI surface filter for impressions and accept that clicks are not coming from Google.
2. **Audit for commodity.** For every top page ask Shepard's question: what here can an AI not replicate — a tool, a dataset, a first-hand test, a decision the reader can act on? If nothing, merge or retire it.
3. **Work the third-party citation graph.** Solís's 84–93% is the number to internalize: list the pages that AI answers actually cite for your category and earn presence there — reviews, comparisons, community, directories.
4. **Keep the crawlers in, keep the hacks out.** Allow citation crawlers (blocking one removes you from that engine entirely), skip chunking tricks and fabricated mentions, and treat llms.txt as free but inert.
5. **Poll, don't check.** Track a fixed prompt set across engines with repeated runs, and report stability. One check per prompt is noise — King's whole essay is about this.

If you want the third and fifth items done for you: the [free AI grader](/ai-grader/) asks ChatGPT, Perplexity and AI Overviews directly, and the [cabinet](/cabinet/) tracks the prompt set weekly with repeated probes and shows which page gets cited instead of yours.

## Sources

- SparkToro × Similarweb, "In 2026, Less than One Third of Google Searches Still Send a Click", 8 June 2026
- SparkToro, "Does Your Website Still Matter in the Zero-Click Era?", 14 August 2026
- SparkToro (Amanda Natividad), "When Google Stops Sending Clicks, What Still Works?", 10 June 2026
- Growth Memo (Kevin Indig), "AI Halftime Report: H1 2026", 27 July 2026; "What makes a good listicle?", 31 August 2026
- Lily Ray, "What We Can Learn from Evolving ChatGPT Fan-Out Queries", 17 August 2026; Search Engine Land interview, 2 September 2026
- Aleyda Solís, "The AI Search Optimization Checklist", updated 20 September 2026; "What It Takes for SaaS Brands to Win in AI Search", 16 July 2026
- iPullRank (Mike King), "There Is No 'Accuracy' in AI Search Tracking – Only Precision", 17 September 2026; zero-click behavior analysis, 30 July 2026
- Zyppy (Cyrus Shepard), "These SEO Strategies Drive 90% of Your AI Visibility", 4 August 2026; "Google AI Ranking Factors", 16 September 2026; "Websites Winning New Google", 9 April 2026
- GSQi (Glenn Gabe), May 2026 core update analysis, 10 June 2026; AI Mode queries via the Search Analytics API, 13 August 2026
- Search Engine Roundtable (Barry Schwartz), 3 June and 1 September 2026
- Google Search Central, "Optimizing your website for generative AI features on Google Search", 15 May 2026; Google I/O 2026 Search updates, 19 May 2026; Search Status Dashboard
- Search Engine Journal, "Google Confirms LLMs.txt Has No Current Implementation", 2 June 2026
- Similarweb, generative AI statistics, 29 July 2026; Ahrefs, llms.txt study, 15 June 2026; Ahrefs, AI Overviews CTR update, 4 February 2026

## FAQ

### What are the main SEO trends in autumn 2026?

Zero-click search as the default (68% of Google searches per SparkToro/Similarweb), Google's official position that optimizing for AI search is still SEO, brand and trust outweighing AI-specific tactics, citations concentrated on third-party sites, a fragmenting AI assistant market, and a measurement crisis where AI surfaces report impressions but not clicks.

### Does Google recommend llms.txt?

No. Google's May 2026 guide on optimizing for generative AI features states that Google Search ignores llms.txt files, and John Mueller called the file 'purely speculative for now'. An Ahrefs study found 97% of llms.txt files received zero requests in May 2026. It does no harm, but it is not a ranking or visibility lever for Google.

### Is GEO different from SEO?

Google says optimizing for generative AI search 'is still SEO'. Practitioners largely agree that classic SEO feeds AI answers — Lily Ray calls AI visibility 'fundamentally an SEO problem' — while Cyrus Shepard notes that the reverse claim, 'GEO is just SEO', is harder to defend because new surfaces such as personal context in Gemini keep appearing.

### How much traffic has zero-click search removed?

It depends on the panel. SparkToro and Similarweb measured 68.01% of US Google searches ending without a click in January–April 2026; iPullRank's analysis of the Datos panel found 46.96% for October 2024 – December 2025. Ahrefs found AI Overviews cut the click-through rate of the first organic position by 58%. Kevin Indig puts publishers' referral loss at roughly 33% year over year.

### Should I block AI crawlers?

The analysts disagree. In a Search Engine Roundtable poll 33% of SEOs said they would block Google's AI features, but Barry Schwartz doubts many will; Google's John Mueller says the biggest hurdle for agentic visibility is sites that block agents; Cyrus Shepard allows that open access is not always the right answer. Blocking a citation crawler removes you from that engine's answers — and Google gives no click data to weigh the trade.