Is SEO dead or has it evolved? What the data shows

With every core update, every AI launch and every viral infographic, the same verdict reappears: SEO is over, everything is different now.

Anyone who lives off organic traffic has learned to distrust the alarm, but this time the question “is SEO dead?” deserves a more careful answer, because some things really did change.

The short answer is no. The useful answer is more interesting: SEO did not die, but three foundations that held the work up were transformed, and much of what circulates as “new SEO” mixes real change with recycled myth.

To tell one from the other, this article leans on the most solid material published on the subject: studies by Ahrefs and Moz covering tens of thousands of pages, peer-reviewed academic research and Google's own public guidelines.

Did old SEO really die? What ended in each era

The old SEO that infographics declare dead had already died more than ten years ago. Mass production of thin content, backlinks bought from any directory and keywords repeated to exhaustion stopped working between 2011 and 2012, when Google launched Panda and Penguin. AI did not kill those practices, it arrived after the funeral.

Caption: anyone asking whether SEO dead is a fact is usually comparing two different eras of the same channel

The timeline helps explain what ended at each moment. Panda, in 2011, took down thin content produced at volume. Penguin, in 2012, turned the low-quality backlink into a toxic liability instead of a shortcut.

Search intent took over right afterwards. With Hummingbird, in 2013, and RankBrain, in 2015, Google began interpreting what the person wants, not only the words they type.

Anyone still describing classic SEO as “fitting in keywords” is talking about a practice buried a decade ago.

E-E-A-T completes the list of myths about old SEO. Since the Medic Update, in 2018, the assessment of experience, expertise, authoritativeness and trust has weighed heavily, especially on health and finance topics. Saying that author credibility “was optional” in old SEO is rewriting history.

That chronology matters for a practical reason: anyone doing good SEO work in 2020 was already applying much of what is sold as news today. What really changed with AI is more specific, and it is measurable.

Why does ranking well no longer guarantee an AI citation?

The deepest change of the AI era is the separation between ranking and being cited. A Moz study of around 40,000 searches showed that 88% of Google AI Mode citations do not appear in the organic top 10 of the same search. The first page and the AI answer became two distinct games.

The same pattern appears outside Google. In the Ahrefs analysis of 3,311 terms, the sources cited by ChatGPT match Google's top 10 in only 10% of cases.

The overlap is so low that treating ranking position as a guarantee of presence in the answers has become a planning error.

The break does not mean ranking lost its value. Organic traffic is still the base of the channel, and AI itself feeds on the same Search index. What ended was the automatic equivalence between position and visibility.

Part of the explanation lies in how AI searches. The systems break the user's question into several sub-questions, a process called query fan-out, and retrieve passages from different pages for each one. Broad topic coverage came to matter as much as the position on an isolated keyword.

Query fan-out rewards whoever built depth. A blog with a complete cluster on a theme has more ways into AI answers than a lone page in first position, and that logic favours digital marketing strategies built on content territories rather than on scattered keywords.

Does schema still work in AI SEO?

Schema markup is the clearest example of a tactic sold as an AI SEO lever with no support in the data. An Ahrefs study followed 1,885 pages that added structured data and compared them with a control group: no platform recorded a citation gain.

The study's numbers are worth recording. Pages that added JSON-LD saw a change of +2.4% in AI Mode and +2.2% in ChatGPT, both indistinguishable from zero, and a 4.6% drop in AI Overviews, which was statistically significant, although the authors themselves do not attribute the drop to schema. Schema is useful technical hygiene for other purposes, but it is not a shortcut to citation.

There is no special structured data nor extra file needed to appear in AI Overviews or AI Mode, and it is Google itself that says so.

Google's recent move reinforces the message, because the FAQ rich result stopped appearing in Search as of 7 May 2026, closing the cycle of a feature that had been restricted since 2023.

The same scepticism applies to llms.txt, the file that promised to “talk to the AIs”. Among the domains that published it, 97% never received a single request for the file in May 2026, according to the logs of 137,000 domains analysed by Ahrefs.

John Mueller, a Google search analyst, summed it up in June 2025: no AI system currently uses llms.txt.

What actually works is less glamorous: extractable structure. AI retrieves passages, so the answer has to sit at the top of the section, the subheadings have to be real questions and each paragraph has to make sense on its own.

Extractable structure is writing and architecture work, like what sustains good indexing on Google, not magic markup.

Video: the on-page SEO checklist that still holds for anyone who thinks SEO dead is settled, on the mkt4edu channel (in Portuguese)

LLM SEO: what makes an AI model cite your brand?

So-called LLM SEO, optimizing to be cited by language models, has academic research behind it. The GEO study, by researchers from Princeton and IIT Delhi, tested nine tactics and found visibility gains of 28% to 41% in the answers, depending on the metric, for the three best ones: citing reliable sources, adding expert quotations and using verifiable statistics.

The study's most valuable detail usually goes unnoticed. Citing sources raised the visibility of pages in fifth position by 115.1%, while the page in first place lost 30.3%. Optimizing for AI benefits those who are not at the top the most, which redistributes the game.

The second strong finding of the AI era involves authority. In the Ahrefs analysis of 75,000 brands, brand mentions on the web correlate from 0.66 to 0.71 with visibility in AI answers, mentions in YouTube videos reach 0.74 and backlinks show a very weak correlation.

The asset that classic link building chased lost its place to the contextual mention. Authority in the AI era is built with digital PR, podcasts and video presence, not with link swaps.

Correlation is not causation, and here comes a hypothesis worth attention. Seer Interactive, after analysing 231,000 LLM responses, argues that what AI thinks about a brand is already written in the training data before the question, and the cited sources appear as support for a pre-formed answer.

If the hypothesis holds, brand building becomes the longest-term SEO work there is.

For anyone producing content, the consequence is direct. Proprietary data, a named expert and a primary source stopped being editorial whim and became the tactics with the largest measured effect on citation, while the new SEO of whoever only rewrites what already exists tends to stay invisible to the models.

Video: the authority techniques that explain why SEO dead is a hasty diagnosis, on the mkt4edu channel (in Portuguese)

What do SEO strategies look like in practice?

SEO strategies were not replaced, they were reorganized into three layers: what classic SEO already required and still holds, what the AI era added and what is emerging as the next frontier. Here is how the three eras line up side by side:

Table: a comparison of the three SEO eras by workstream, the fullest answer for anyone asking whether SEO dead is true

The AI era is where the work happens today, and it does not cancel classic SEO. The technical base still conditions everything: a slow or badly indexed page enters neither the ranking nor the AI answers, and Core Web Vitals problems keep sinking the result before any discussion about citation.

Measurement is the front that demands the most immediate adaptation. An AI model's answer varies between runs and between users, so the fixed position gave way to frequency: how often the brand appears when the same question is asked many times.

Citation tracking does not exist in Google Analytics and tends to require a dedicated tool.

Anyone who does not know where to start solves it with a diagnosis. An SEO audit reveals what already works, what broke and where the brand already appears in AI answers, and a mature SEO service today runs all three layers at the same time.

Which trends will shape the new SEO in the coming years?

Four trends should shape the new SEO in the coming years: ads inside AI answers, paid access to content, commerce run by agents and search with memory. None is stage futurology, each has a concrete move already announced by the platforms, and it is worth separating what is settled fact from what is a projection on signals.

Monetizing AI answers is the most advanced trend. Google is testing sponsored formats inside AI Mode and has expanded the Direct Offers pilot, launched in January 2026, with retailers such as Chewy, Gap and L'Oréal.

The scale of the investment confirms this is not an experiment. Alphabet projected an infrastructure investment of US$175 billion to US$185 billion for 2026, against US$91.4 billion spent on property and equipment in 2025, with AI as the priority. Organic and paid are likely to compete for the same answer space.

Access to content became a negotiable commodity. Cloudflare created Pay per Crawl, in which an AI bot without a payment agreement is turned away, and since July 2025 it has changed the platform default to block AI crawlers that do not pay for content.

Deciding who may read your content, free of charge or for a fee, is ceasing to be a technical setting and becoming a business decision.

Commerce run by AI agents had its first instructive retreat. OpenAI launched Instant Checkout in September 2025, the purchase completed inside ChatGPT, and in March 2026 moved the function to retailer apps, which take the buyer to the store's own site.

The rails keep being built even after the retreat. Google's agent payments protocol brings together more than 60 organizations, among them Mastercard, PayPal and American Express.

The fourth trend is the most structural, and here the projection is ours: search with memory. Assistants that keep user context tend to personalize every answer, which should retire traditional rank tracking for good and turn monitoring into statistical sampling by persona and by platform.

Whoever starts measuring AI presence now arrives at that scenario with a history, and a visibility history is the kind of advantage you cannot buy later.

Video: the paid versus organic debate that returns whenever someone claims SEO dead is a fact, on the mkt4edu channel (in Portuguese)

Is SEO dead: frequently asked questions

Where should you start your SEO for the AI era?

The starting point is not redoing everything, it is finding out what already works and what broke. SEO did not die, but the yardstick moved: the technical base is still eliminatory, content structure decides who is cited and brand authority decides who is remembered. Without a diagnosis, any move is a bet.

The next step already has a guide. The layers of AI SEO show, one by one, what to adjust in the technical base, in the content and in brand authority so your company enters the answers, and they are the natural continuation of what this article mapped.

If the doubt that brought you here was “is SEO dead?”, the best answer is practical: audit your site, measure your presence in AI answers and decide with data. The channel that survived Panda, Penguin and mobile did not die with AI, it only became more demanding of whoever operates it.