GeoCited
Book a call

Is SEO Dead? The Future of SEO When Buyers Ask AI

SEO is not dead. The future of SEO is a job with the same inputs and a different scoreboard: you still earn pages a retrieval system can read, trust and quote, but the system doing the quoting is now an assistant rather than a results page. Rankings did not stop mattering. They stopped being the last step before a buyer sees you.

Lourdes Paul Agilan, founder of GeoCited

Founder, GeoCited · Published · 7 min

The future of SEO when buyers ask ChatGPT instead of Google: what the retrieval data actually shows, which numbers hold up, and what changes in the daily work.

On this page

Note on verification: GeoCited's own 40-prompt citation study is our agency's research, published with its raw data but not independently reviewed, so treat it as vendor research. Pew's click-through data comes from browsing behaviour in March 2025, which is older than everything else here. The SparkToro and Datos share figures are desktop clickstream only, so mobile and in-app search are missing. My view on where inference runs next is opinion, not a forecast anyone has measured.

What people are really asking when they ask "is seo dead"

Nobody asking this wants a definition. They want to know whether the budget line survives the year.

I get the question from founders who watched non-branded clicks slide while their pipeline held steady, and from marketers whose CEO used ChatGPT once and came back convinced Google is over.

Both are describing the same thing from different angles. Demand did not move. The layer that decides who gets named moved.

So the honest version of the question is not whether SEO is dead. It is which parts of the work still pay, and who is now reading the pages you write.

The retrieval layer changed hands, and it still reads Google

Here is the finding that settles most of the argument for me.

RESONEO analysed 1,249 ChatGPT answers that involved retrieval, published in July 2026 and updated in August. One retrieved URL in three already appeared on Google's first page. Only one in twenty appeared on Bing's.

Read that twice. The assistant people describe as Google's replacement is, a third of the time, handing back pages Google already ranked.

That is not a coincidence, and it is not permanent either. ChatGPT pulls results from its own index and from scraped Google results, and the mix shifts by account tier and by model version. I walked through that pipeline in detail in how ChatGPT searches now.

The point stands regardless of the mix. Your Google position is an input to the AI answer, not a casualty of it.

RESONEO's second number matters more for the work itself. A page ChatGPT actually opened ended up cited 74% of the time. A page that sat in the grounding set without being opened was cited 7% of the time.

Being retrieved is not the win. Being opened is.

The numbers behind "is seo dead with ai"

The pessimists are not making things up. Two independent findings deserve to be reported straight, because they point in opposite directions.

Pew Research Center tracked the browsing of 900 US adults in March 2025. When an AI summary appeared, people clicked a traditional search result in 8% of visits. Without a summary, 15%. They clicked a link inside the summary itself in 1% of visits.

That is a real loss of clicks, from an independent research organisation with no product to sell. It is also more than a year old now, and AI summaries appeared on only 18% of searches in that sample.

Against it, scale. SparkToro's March 2026 analysis of Datos desktop clickstream data across 41 search destinations put Google at 73.7% of US desktop searches in the last quarter of 2025, with AI tools combined at about 3.2% and ChatGPT alone under 3%. Amazon, Bing and YouTube each took more desktop search activity than ChatGPT did. Google lost about 3.5 percentage points of share between the first and last quarter of the year.

SparkToro sells software and Datos is owned by Semrush, so treat those as vendor research. Desktop-only data also flatters Google, since a lot of assistant use is in apps.

Put both together and you get something less dramatic than either headline. A large channel is leaking clicks slowly while a small channel grows. That is a reallocation problem, not a funeral.

The future of SEO is the same craft with a stranger scoreboard

What genuinely changed is measurement, and that is where I see most teams still working with the wrong instrument.

There is no ranking inside ChatGPT. Not a hidden one, not a proprietary one, not one an agency can chart for you.

SparkToro tested this properly in January 2026: 600 volunteers, 2,961 runs across ChatGPT, Claude and Google's AI. Asked the same question repeatedly, there was less than a 1 in 100 chance that any two answers returned the same list of brands.

So anyone selling you a guaranteed AI placement is selling you a coin flip with a dashboard attached. One screenshot proves nothing. A distribution across repeated runs is the only honest reading.

Prompt research is the other new skill. Keywords were phrases people typed. Prompts are questions people ask in full sentences, and the page type that wins depends on the question shape.

In our own study in August 2026, 40 buying-intent prompts run on ChatGPT and Google across 490 cited URLs, comparison pages led "X vs Y" questions at 56.7% and pricing pages led cost questions at 53.4%. Product pages took 37.7% of ChatGPT's citations. That is our own work on one query set, so hold it loosely, but it matches what I see on client accounts: the intent decides the asset.

The technical layer, by contrast, is old news wearing new clothes. Search Engine Land reported in August 2026 that ChatGPT's crawler does not execute JavaScript, that pages over 4 MB are rejected outright rather than truncated, and that in instant mode the roughly 200 characters after your H1 are the only body text the model sees.

Server-rendered HTML. Reasonable page weight. A direct answer high on the page. Every one of those was good SEO advice in 2015.

Access is the one genuinely new gate. OpenAI's publisher FAQ is blunt about it: if you block OAI-SearchBot, your content is not included in ChatGPT's summaries and snippets. GPTBot is the separate training crawler. Referral clicks arrive tagged with utm_source=chatgpt.com, which is the cheapest measurement you will set up all quarter.

I have seen more visibility lost to a firewall rule than to a competitor's content.

What I would change this quarter, and what I would leave alone

Leave alone: technical hygiene, internal linking, topical depth, and earning mentions on sites the models already read. That work now pays into two channels instead of one.

Change three things.

Measure differently. Run your buying questions repeatedly across engines and record a share of answers, not a position. Our AI visibility audit is built around that, and the useful signal shows up on the second and third run, never the first.

Build for question shapes. If comparison and pricing questions dominate your category, a thin pricing page is a visibility problem, not just a conversion one.

Check your access logs before you buy anything. Confirm the retrieval crawlers can reach you, then confirm which of your pages get opened rather than merely listed.

If you want that handled as an ongoing programme rather than a one-off check, that is what our AI SEO services cover: prompt discovery, citation gap analysis, and weekly tracking across the engines.

Where I think this goes next

This part is my opinion, and I want it labelled as such.

I expect a meaningful share of assistant queries to be answered by models running locally, on phones and laptops, within a few years. If that happens, retrieval gets cheaper and more frequent, and the fetch that decides your visibility may not touch your server at all.

I cannot prove that, and nobody has measured it. I mention it because it is the scenario that would actually break current measurement, rather than the one everyone is worried about.

The scenario people fear, Google going away, is not the one the data supports.

One question for you

Pull your last 20 closed-won deals: how many of those buyers found you through a ranked link, and how many arrived already having asked an assistant about your category?

Sources

Frequently asked

Is SEO dead in 2026?

No. Google still took 73.7% of US desktop searches across 41 search destinations in the last quarter of 2025, per SparkToro's analysis of Datos clickstream data, while AI tools combined took about 3.2%. Google lost about 3.5 points of share between the first and last quarter of that year, which is a reallocation of effort rather than the end of the channel.

Do Google rankings still matter if buyers ask AI?

Yes. RESONEO found that one in three URLs ChatGPT retrieved in its July 2026 sample already sat on Google's first page. Ranking well is one of the ways you become eligible to be quoted.

Can an agency guarantee my brand appears in ChatGPT?

No, and I would walk away from anyone who says otherwise. SparkToro's January 2026 study of 2,961 runs found less than a 1 in 100 chance that two answers to the same question list the same brands. There is no ranking to buy or hold.

What is actually different about optimising for AI answers?

Measurement and prompt research. The tactics overlap heavily with good SEO: crawlable HTML, clear answers near the top, depth, and third-party mentions. What is new is tracking a share of answers across repeated runs, and researching questions rather than keyword strings.

Will blocking AI crawlers protect my content?

It will also remove you from the answers. OpenAI's publisher FAQ states that content is included in ChatGPT summaries and snippets only if OAI-SearchBot is not blocked, while GPTBot handles training separately. That separation lets you opt out of training without opting out of visibility.

Written by

Lourdes Paul Agilan, founder of GeoCited

Founder, GeoCited

Lourdes Paul Agilan runs GeoCited, a generative engine optimization agency. He works with B2B companies on how AI assistants describe and recommend them. He writes about that work here, and shares the experiments behind it. The tests that worked, and the tests that did not, written up the same way.

LinkedIn

Related reading

Find out what AI says about you.

We run your buyers' real questions against the major answer engines and write down every citation.

Book a callSee the work