Does SEO still matter if your customers are asking ChatGPT?

It does — but not for the reason it used to. The job of ranking has quietly changed from winning a click to getting into the answer. That distinction decides what is worth publishing.

What actually changed

The numbers being passed around are dramatic, and worth reading with one eyebrow raised.

Reported March 2026 figures put Google at roughly 90% of conventional search, with ChatGPT holding about two-thirds of AI chatbot sessions. AI Overviews now appear on around 18% of all Google searches, and far more of the long, specific ones. Where an AI Overview appears, organic click-through has been reported to fall from about 1.76% to 0.61% — roughly a 60% drop. Overall zero-click rates sit near 43%.

Two honest caveats. Most of these statistics are published by companies selling AI-search services, ourselves included in spirit — we sell this work. And the field is eighteen months old, so anyone offering you certainty is selling something. Directionally, though, the pattern is consistent everywhere: fewer people click, and the ones who do are further along.

That second half matters more than the first, and gets quoted less. AI-referred visitors have been reported converting at around 14% against 2.8% for traditional organic. Whatever the precise figure, the mechanism is obvious: someone who arrives after an assistant has already answered their question and recommended you is a warmer prospect than someone who clicked a blue link to find out what a thing is.

Why ranking matters even when nobody clicks

Here is the part that gets lost in "SEO is dead" arguments.

When you ask ChatGPT or Google's AI Mode about something current, it does not recall an answer from training. It runs searches, retrieves pages, reads them, and writes a summary of what it found. That retrieval step uses search infrastructure.

If you do not rank, you are not in the pile the assistant reads. You cannot be cited from outside the room.

So the work has not changed as much as the vocabulary suggests. Being findable, being clear, being specific and being demonstrably credible were always the fundamentals. They now decide whether a machine can quote you, rather than whether a human clicks you.

The practical consequence is that SEO and so-called GEO are not competing budgets. Content that ranks well is disproportionately the content that gets cited. One is the prerequisite for the other.

What genuinely stopped working

Something real did die, and pretending otherwise is how agencies keep selling work that no longer pays.

Content whose entire value was answering a simple question — a definition, a date, a conversion, a one-step how-to — has lost its traffic permanently. The assistant resolves that in the answer box. Nobody needs your page to learn what a marketing funnel is.

Which means the following are now mostly wasted effort:

• "What is X" explainers on topics already covered a thousand times
• Volume blogging aimed at keyword coverage rather than a reader
• Thin listicles that restate what is already on page one
• Anything an assistant can fully answer in two sentences

If a machine can answer the question completely without you, the page was never really yours to own.

What survived, and got more valuable

The content that held up is the content a summary cannot replace:

• Comparisons and decisions. "Should we do A or B, given our situation" needs judgment, trade-offs and context — and readers still click through to see the reasoning.
• First-hand experience. What actually happened when you did the thing. An assistant cannot synthesise something nobody has written down.
• Original numbers. Your own data, however small, is the most citable thing you can publish.
• Specificity. "Marketing for Indian business schools" is winnable in a way "digital marketing" is not — for both search engines and assistants.
• A clearly described business. When someone asks an assistant to recommend a firm, it answers from what the web consistently says you are. That is an entity problem, not a blogging one.

Your traffic report is about to mislead you

This is the trap we see most often. A business does good work, gets cited in AI answers, and sees sessions fall. They conclude SEO is failing and cut it — at the exact moment it started working differently.

If your only measure is organic sessions, you will misread the situation. Worth watching instead:

• Branded search volume. If more people search your name directly, something is introducing you to them. Often that something is an assistant.
• Direct and referral traffic quality. Fewer visitors converting at a higher rate is a good outcome, not a bad one.
• Whether you appear at all. Ask ChatGPT, Gemini and Perplexity the questions your customers would ask. Do it monthly. It is crude, free, and more informative than most dashboards.
• Enquiry quality. If people arrive already knowing what you do, the machines are describing you accurately.

What a small business should actually do

Nothing exotic. In order:

1. Make sure your pages can be read. Plenty of modern sites render entirely in JavaScript. Google copes; several AI crawlers do not run scripts at all and see an empty page. Worth checking before anything else.
2. Answer the question in the first paragraph. Assistants lift the top of a page. Building to a conclusion at the bottom is a habit from a different era.
3. Say who wrote it, and why they would know. A named author with real credentials is a credibility signal for both readers and machines.
4. Describe your business identically everywhere. Site, listings, profiles. Inconsistency makes you harder to summarise.
5. Publish fewer, better pages. Three pieces with something genuinely in them will do more than thirty that restate the obvious.

The uncomfortable part

A lot of what is currently sold as "GEO" or "AEO services" is SEO with a new invoice. Structured data, clear headings, credible authorship and content worth citing were best practice long before ChatGPT existed.

There are genuinely new things to do — tracking whether assistants mention you, structuring content so answers can be extracted cleanly, thinking about your business as an entity rather than a set of pages. But if someone offers you an AI-search package and cannot explain what is different about it in plain words, it is probably the same work at a higher price.

Ask them. It is a fair question, and the answer tells you a lot.

Common questions

Is SEO dead in 2026?

No, but its job changed. AI assistants ground their answers in search results, so a page that does not rank is usually not in the pool an assistant can draw from. Ranking is now how you become eligible to be cited rather than how you win a click. What has genuinely stopped working is content whose only purpose was to capture a click for a simple factual question.

How do I get my brand mentioned by ChatGPT?

Be findable, be specific, and be consistently described. Publish pages that answer real questions with the conclusion in the first paragraph, use clear headings and structured data, and make sure your business is described the same way across your site, your listings and your profiles. Assistants assemble answers about a business from what the web consistently says about it, so consistency matters as much as volume.

What is the difference between SEO, GEO and AEO?

SEO is optimising to rank in a list of search results. AEO, or answer engine optimisation, is structuring content so a direct answer can be lifted from it. GEO, or generative engine optimisation, is being cited inside an AI-generated response. In practice they overlap heavily: pages that rank well are the ones that tend to get cited, so GEO and AEO are best understood as extensions of good SEO rather than replacements for it.

Where these figures came from

Market share, AI Overview prevalence and click-through figures are reported March 2026 data from Stackmatix. Findings on which content types held or lost traffic are from Contently's analysis of 2026 AI Overview traffic data. Both are commercially interested parties in a field with little independent measurement, which is why this article gives directions rather than precise promises.