Is SEO dead in 2026? The honest answer in numbers

No, SEO is not dead in 2026. Google's own guide to AI search, last updated in July 2026, asks "Is SEO still relevant for generative AI search?" and answers "In short, yes!" Google still sent 91.39 percent of UK search engine referrals in September 2026, by StatCounter's count. But the answer is not the same for everyone. Publisher SEO is dying. Commercial SEO is intact but harder. GEO is mostly a rebrand. The rest of this post is the data behind each claim, with the main figures re-checked in October 2026, plus what the work actually looks like now if you take it seriously.

I have been doing SEO professionally since 2007. I founded SEO Moves in 2014 and still run it. The "is SEO dead" question has been asked every year for at least the last decade. This year, for the first time, the answer is meaningfully nuanced.

What is genuinely, measurably dying#

Let's start with the parts you cannot wave away.

A Reuters Institute report using Chartbeat data from more than 2,500 sites found Google search traffic down 33 percent globally between November 2024 and November 2025. US publishers fared worse at 38 percent. Google Discover dropped 21 percent globally, 29 percent in the US. Publishers surveyed for the same report expect search traffic to fall a further 43 percent over the next three years. It then got worse. In October 2026 the Reuters Institute reported Chartbeat's newest figures: Google Search referrals to publishers fell 40 percent from July 2025 to July 2026, twice the fall of the year before. You cannot dismiss that as small-sample noise.

The vertical breakdown is worse where you look closely. A Growtika study by Yuval Halevi, published in February 2026 and built on Ahrefs' US traffic estimates, tracked the US Google traffic of 10 major tech publications. Between their 2024 peak and January 2026 they lost a combined 58 percent of US organic search traffic - 65 million monthly visits gone. The Verge alone is down 85 percent. The same study checked two big publishers outside tech: NerdWallet lost 73 percent and Healthline 50 percent. Stack Overflow's volume of new questions was down 75 percent from its 2017 peak by December 2024, and 60 percent year-on-year. Chegg sued Google over AI Overviews. On 30 September 2026 a US judge dismissed that case, and Penske Media's, ruling that the claims "fail to get out of the starting gate".

The Guardian called the recipe space an "extinction event" for independent food bloggers in 2025. 30 to 50 percent drops are routine in that vertical now. Personal finance got hit at NerdWallet scale across the board. Lifestyle, weather, horoscopes, TV guides - anything where the user wants a quick fact rather than a deeper interaction - all show similar patterns.

Fewer searches end in a click at all. SparkToro and Similarweb found 68.01 percent of US Google searches ended without a click in the first four months of 2026, up from 60.45 percent in 2024. A follow-up in June 2026 put the UK at 69.5 percent, the highest of the six countries measured.

This is the publisher SEO model dying. It is real, it is measured by independent third parties, and it is not coming back. Press Gazette covered it in January 2026 under the headline "Global publisher Google traffic dropped by a third in 2025".

If your business is "we publish 50 articles a month and monetise via display ads," then yes - your SEO is dead. The question is whether you call that SEO dying or the ad-funded publishing model dying. They are the same thing in practice but they are not the same thing analytically.

What is intact but harder#

Now the other side. The same 12 months that broke publisher SEO did relatively little to commercial-intent SEO.

AI Overviews are not on every search. BrightEdge's one-year report, published in February 2026, found them on about 48 percent of the queries it tracks, with about 52 percent still showing none. My read is that Google has less reason to AI-summarise a query where the user is about to click through to a transaction, because commercial clicks are what its ads monetise. Google told investors in July 2026 that AI Mode is driving "an incremental increase in Search queries overall" and that it sends "billions of clicks to websites every week through AI features". Those are Google's claims about its own product, with no counts attached.

Seer Interactive measured organic CTR on AI Overview queries falling 61 percent, from 1.76 percent in June 2024 to 0.61 percent in September 2025. Its April 2026 update used a bigger, different dataset: 5.47 million queries across 53 brands. There, CTR on AI Overview queries climbed from 1.3 percent in December 2025 to 2.4 percent in February 2026. In Seer's words, "The decline didn't level, but instead bounced back a little". Seer also warns against forecasting a recovery from two months of data. Other studies are less kind. Ahrefs found the top result's CTR is 58 percent lower when an AI Overview shows (December 2025 data). When AI Overviews launched in France, Ahrefs tracked 963 domains: CTR fell 3.4 percent overall in nine days, and 23.1 percent for the most exposed.

Ahrefs ran a 76,000-site analysis in February 2026 and found Google sent 190 times more traffic to websites than ChatGPT did: 39.98 percent of traffic to the sites it tracked, against 0.21 percent. Its live tracker has moved on. Updated on 13 October 2026 and covering 115,254 sites, it shows Google at 23.96 percent of tracked traffic in September 2026, ChatGPT at 0.34 percent and all AI assistants together at 0.43 percent. By my arithmetic that is about 70 times more traffic from Google than ChatGPT. The two figures use different bases, so do not read 190 to 70 as a precise trend. Semrush's data points the same way: AI traffic grew 66 percent in 2025 but was still less than 0.15 percent of total visits, and ChatGPT's outbound referrals grew 206 percent from January 2025 to January 2026 (US clickstream). BrightEdge found ChatGPT referral traffic up 101 percent from January to August 2026. That is real momentum, but the absolute volume is still small enough that "AI search replaced Google" is not yet a true statement.

There is one important wrinkle. AI-referred visitors convert dramatically better than Google organic visitors in the studies that exist. Seer Interactive found ChatGPT visitors converted at 15.9 percent versus 1.76 percent for Google organic, for a single client. Ahrefs measured a 23x higher conversion rate on its own site: AI search sent 0.5 percent of its traffic and 12.1 percent of its signups. Semrush found the average AI search visitor worth 4.4 times the average traditional organic search visitor, based on conversion rate. AI search is a small but extremely high-quality traffic source.

Commercial SEO in 2026 is harder, costs in my estimate about 20-30 percent more to do properly (because GEO work has to be added), and pays back if you do it right. It is not dead.

What structurally changed under the hood#

Three changes that matter more than people realise.

One. Backlinks lost their throne#

Google removed the line "Google uses links as a factor in determining the relevancy of web pages" from its spam policy in the second half of 2024. It is on the version of the page last updated 31 July 2024 and gone from the version last updated 19 November 2024. That is not a minor edit, that is a load-bearing sentence being deleted. Gary Illyes from Google said in 2023 that links were no longer a top-three ranking factor. John Mueller, Duy Nguyen and other Search Relations staff have echoed it since. Russell Welch wrote in Search Engine Land in December 2025 that Google and Bing now weigh unlinked brand mentions and entity references alongside backlinks, and LLM-driven surfaces amplify this because they parse entity prominence, not anchor text.

Anyone selling you "100 backlinks per month" packages in 2026 is selling you a 2014 product. The contemporary version of that work is digital PR that earns brand mentions, podcast appearances that create transcripts, original research that gets cited.

Two. Brand became the dominant signal#

The Google API leak in May 2024 confirmed something the experienced SEO community had suspected for years. NavBoost - the system that ranks pages partly on aggregated user click behaviour from Chrome - uses up to 13 months of click data. Pages that get more clicks for a query rank higher. Brands that users already recognise get clicked more. This creates a flywheel. Big known brands win because users click them, which trains the algorithm to show them more, which earns more clicks.

This is also why Helpful Content System updates appeared to disproportionately hit independent publishers and reward established media brands. HouseFresh, a small but rigorous review site that does hands-on testing of air purifiers, lost ~90 percent of its traffic to the September 2023 HCU. Gisele Navarro who runs it gave congressional testimony about it. HouseFresh substantially recovered in the June 2025 core update - reportedly without changing much - because Google's classifier finally caught up to the fact they were the actual experts.

The lesson is brutal. Whether your content ranks now depends on whether your brand is known. Pure content quality with no brand signal underneath gets devalued. This is not in any official Google document. It is observable across hundreds of recovery stories.

Three. Query fan-out changed what "rank for a keyword" means#

AI Overviews and AI Mode do not rank one page for one query. Google's own guide describes query fan-out as "a set of concurrent, related queries generated by the model". The system retrieves results for all of them and synthesises one answer from many pages. Mike King at iPullRank built the Qforia query fan-out simulator to model this. BrightEdge measured in February 2026 that only about 17 percent of AIO citations come from the organic top 10 for the same query, and that figure has barely moved.

This is the biggest structural shift and the one most SEOs are still ignoring. Optimising your page to rank #1 for "best CRM for nonprofits" tells you almost nothing about whether AI Mode will cite you when someone asks "what CRM should our food bank use that integrates with our donor management system." The mental model has to shift from "what keyword does this page target" to "what cluster of questions could this page answer."

The GEO question#

Here is the uncomfortable part. GEO - generative engine optimisation - is mostly a rebrand of SEO work that good SEOs have done for years.

The actual mechanics:

• Entity optimisation (sameAs links, Wikidata, consistent NAP) - SEO has done this since 2014
• Schema markup - SEO since 2011
• EEAT signals (named authors, credentials, trust) - SEO since 2018
• Citation seeding on third-party platforms - digital PR since 2010
• Original data and research as a citation lever - content marketing since forever

The bits that genuinely add to the SEO toolkit:

• Query fan-out resilience - making sure your page answers the question someone would actually ask, not just the keyword Google indexes
• Sections that make sense on their own - good writing more than a new discipline. For Google, there is "no requirement to break your content into tiny pieces", in the words of its own AI search guide
• Recency and named, verifiable authors - both are widely sold as AI citation levers. I have not found a primary study that measures how much they matter, so treat any precise percentage you are quoted as unproven

That is a real but small addition to the toolkit. Google now says the same thing more bluntly. Its AI search guide calls optimising for generative AI search "still SEO", and says that for Google Search you can ignore tactics like chunking content, unnecessary AI text files such as llms.txt, and chasing inauthentic mentions. Whether you call it GEO or call it "the next iteration of SEO" matters less than whether you do the work.

What is grift is the GEO consultant selling six-figure retainers to clients who never did proper SEO in the first place. On 15 May 2026 Google clarified that its spam policies cover generative AI responses too: spam now includes "attempting to manipulate generative AI responses in Google Search". That was a documentation change, not a new algorithm, and the policy does not name specific GEO tactics. In my view the most popular ones being sold right now - self-promoting your client into AI-generated comparison lists, entity stuffing, scaled "best [X]" pages - are the obvious targets. Google has run spam updates since, starting 24 June, 18 August and 24 September 2026.

Amos Weiskopf wrote a piece in December 2025 called "GEO is Premature Optimization." David McSweeney wrote "The Great GEO Grift" in October 2025. The sceptics are mostly right. Google has closed part of the gap: by 31 August 2026 Search Console's Generative AI performance report had reached every site, showing impressions in AI Overviews and AI Mode. I have not found an equivalent for ChatGPT or Perplexity, and there is still no validated causal model for LLM citations. Anyone selling certainty about LLM rankings in 2026 is selling you confidence theatre.

Lorelight - one of the higher-profile GEO measurement platforms - shut down on 31 October 2025. The founder's announcement was a clear-eyed admission: "There's no secret formula. No hack. No 'GEO strategy' separate from just doing good marketing." Read that post if you are about to spend money on a "GEO platform."

What the UK market actually looks like#

The UK picture has specific shapes the US-centric coverage misses.

The CMA designated Google Search as having Strategic Market Status under the Digital Markets, Competition and Consumers Act 2024 on 10 October 2025. AI Overviews and AI Mode are explicitly in scope. Combined with the parallel mobile platform SMS designation on 22 October 2025, the CMA now has activity-level granularity over Google in the UK that Brussels does not quite have under the Digital Markets Act. It has used it. The CMA published its conduct requirements consultation on 28 January 2026. On 3 June 2026 it imposed a publisher conduct requirement: Google must give publishers effective controls over how their content is used in its generative AI features, and clear metrics on how users engage with that content. Fair ranking and data portability requirements followed on 17 June 2026, and a consultation on user choice opened on 23 September 2026. By 31 August 2026 Google had also rolled out a Search Console toggle, to every site worldwide, that lets you opt out of AI Overviews and AI Mode. Google says the toggle is not used as a ranking signal outside those features.

UK AI Mode launched on 28 July 2025. The PPA's evidence to the CMA showed one lifestyle publisher's CTR on one query falling from 5.1 percent to 0.6 percent over a year. Daily Mail reported around 50 percent referral drops. Sistrix found position-one CTR drops from 27 percent to 11 percent when an AI Overview shows, a fall of almost 60 percent, in a German study of over 100 million keywords published in February 2026. Its UK newsbrand data for January to April 2026 found health the vertical most exposed to AI Overviews, followed by tech, travel, lifestyle and money.

The UK agency market is consolidating fast. Brave Bison acquired Builtvisible in March 2025 for up to £3.5 million, against £4.1 million of revenue - under 1x revenue, which is the multiple distressed agencies trade at. Atlas SEO in Norwich proposed closing in February 2026, putting more than 50 jobs at risk, less than two years after a reported £2.71 million acquisition. Stickyeyes (Leeds) was absorbed into Pixis.ai. Four Magazine and Digital Applied are both predicting around 30 percent of generalist UK agencies will fold or merge by end of 2026.

But - and this is the contradiction worth holding - UK searches for "SEO agency" are up 61 percent year-on-year (Koozai April 2025 report). "PPC agency" up 115 percent. Demand from buyers is up. Supply of cheap commodity SEO is collapsing. The market is restructuring, not dying.

ITJobsWatch no longer has enough recent vacancies to give a median day rate for an SEO consultant (checked October 2026). Across all UK contracts that mention SEO, the median day rate in the six months to 14 October 2026 was £413, up 37.5 percent on £300 a year earlier. That rests on only 43 quoted rates, so read it as a sign, not a trend. The senior independent end - day rates of £400-£700 - is doing fine because the work that band does (strategic advisory for marketing teams that already know they need help) is the work the AI revolution has actually increased demand for.

What the work looks like now#

Practical, not theoretical. This is what I am actually doing for UK clients in 2026 that I was not doing in 2022.

Entity audit first. Before any content work, audit the client's entity footprint. Wikidata entry exists? Google Knowledge Panel claimed? sameAs links from LinkedIn, Companies House, Crunchbase, GitHub if they have one? Named authors with verifiable credentials? This audit takes a day and finds the work the next six months should focus on. Until 2023 most clients had a passable footprint by accident. Now most do not.

Original-data content as a structural priority. Stop writing "10 ways to..." aggregations of other people's ideas. Find one thing you have proprietary data on. Publish it. Become the source, so the answer has to cite you rather than someone summarising you.

Schema that matches the visible page. Person, Organization, Service, LocalBusiness, SoftwareApplication, Article with author and dateModified. Google's AI search guide says structured data isn't required for its AI features and there is no special markup to add, though it still helps with rich results. Google also stopped showing FAQ rich results on 7 May 2026. I still add schema because it states who you are and what you sell in a form machines can read without guessing. Pages with it that lie are spam. Pages with it that match the visible content are the only legitimate option.

Citation seeding on the platforms AI engines pull from. In the citation studies I had read by May 2026, Reddit was the single most-cited domain across ChatGPT, Perplexity and Google AI Overviews. Since then Ahrefs measured ChatGPT's Reddit citations in the US falling 88 percent between March and September 2026. LinkedIn for B2B exclusively. YouTube transcripts for procedural content. G2 and Capterra for B2B SaaS. This is digital PR work, but pointed at the platforms that feed AI rather than at TechCrunch.

Defensive technical SEO, not chase-the-trend. Core Web Vitals at the green threshold. Crawl efficiency. No render-blocking client-side JavaScript on money pages. Internal linking that puts authority on commercial templates. None of this has changed since 2020. All of it still matters.

Conscious decisions about GEO work versus SEO work. Some changes help one and hurt the other. Adding a spoiler-style meta description that summarises the page hurts Google CTR but helps ChatGPT extraction. Inserting question-first H2s helps both. The choice between them is now strategic, not technical.

This is mostly the same craft. The proportions have shifted. The mental model needed an update.

What changed between May and October 2026#

None of this changes the verdict. It sharpens it.

• Google's AI features got bigger. At I/O on 19 May 2026 Google said AI Overviews have over 2.5 billion monthly active users and AI Mode over 1 billion. You can now ask a follow-up from an AI Overview straight into AI Mode, worldwide.
• ChatGPT grew too. OpenAI says it reaches 1.2 billion people a week (October 2026), up from more than 900 million in March 2026. Those are OpenAI's own figures.
• You can now see and control your exposure in Google. Search Console's Generative AI performance report and the AI features opt-out reached every site worldwide by 31 August 2026.
• Google kept updating rankings. A core update started 21 May 2026, and spam updates started 24 June, 18 August and 24 September 2026. No core update has been listed since May.

Honest verdict#

Publisher SEO funded by ad pageviews on informational content: dead or dying. If your business is here, treat 2026 as an exit ramp.

Commercial SEO for transactional, lead-gen and local businesses: intact, more expensive, harder, still the highest-quality traffic source you can buy. In my estimate, budget needs to go up about 20-30 percent to cover GEO work properly. Returns should justify it.

GEO as a separate discipline: mostly a rebrand, and Google's own AI search guide now calls it "still SEO". The genuinely new mechanics (query fan-out, citation seeding) sit inside the SEO toolset, not outside it. Be sceptical of anyone selling it as a separate practice with a separate retainer.

The "SEO is dead" people are right about a real, large, measurable part of the market. The "SEO is alive" people are right about a different real, large, measurable part of the market. Both sides are missing the structural restructuring that makes the question wrong-shaped.

If you want a calibrated, honest read on whether your specific situation needs more, less or different SEO work, the SEO pricing calculator gives you a band in two minutes. If you want an actual second opinion from a senior independent who is not selling you an agency retainer, the contact form is the way in.

Just stop asking the wrong question.