Why SEO still matters when AI answers questions

The old argument for SEO stopped being true in 2025

For fifteen years the case for SEO was one sentence: organic results get most of the clicks, ads get the rest, so build the thing that doesn't charge per click. It was a decent argument and it's now half wrong.

AI Overviews and AI Mode answer a large share of questions on the results page itself. The user reads the summary and stops. That isn't a prediction or an agency scare story — it's measurable, and the measurement is worse than most SEO pitches admit.

Which means the honest case for SEO in 2026 has to be rebuilt on different ground. Here's that ground, including the parts that don't favour us.

What the click data actually shows

The most credible public measurement comes from the Pew Research Center, published in July 2025. Rather than scraping search results, Pew analysed the real browsing behaviour of a panel of around 900 US adults who agreed to share their data, covering their Google searches during March 2025.

Two findings matter. Users clicked through to a search result on roughly 8% of visits where an AI summary appeared, compared with about 15% of visits where no summary appeared. And roughly 1% of visits with a summary produced a click on a link inside the summary itself.

Read those numbers carefully, because they're routinely overstated. It's a small US panel over one month, it measures visits rather than queries, and it doesn't separate commercial from informational intent in a way you can plan a budget around. Plenty of vendors publish bigger, scarier numbers from scraped SERPs. None of those are audited, and they disagree with each other. The Pew figures are the ones we're willing to put our name to, and even they only support one conclusion: fewer clicks per search than before.

Why AI answers still need you to rank

Here's the part that saves the argument. Google's AI Overviews are not a separate model reading the open web fresh for each question. They're generated from retrieval against Google's existing search index. The summary is written from pages the index already holds and rates as good candidates, and the citations come from that same retrieval set.

On top of that, a question gets decomposed into related sub-questions, each of which retrieves its own candidates. That's why pages get cited for questions they never explicitly targeted — they happened to be the strongest candidate for one sub-question inside a bigger one.

The practical consequence is unglamorous and useful: the work that gets you cited is mostly the work that got you ranking. Be indexed. Be a strong candidate for specific sub-questions rather than only the head term. Put the answer in a short, self-contained paragraph a machine can lift without needing the three paragraphs around it. Keep numbers and specifics in text, not in an image or a PDF.

One honesty note. Google has never published how sources are selected, citation sets shift between sessions and users, and there is no verified method to guarantee a citation. Anyone selling you one is selling a guess with an invoice attached. What you can do is make yourself a strong candidate, which is what SEO already was.

The cost argument, which got stronger

Every acquisition channel except organic search charges rent. Ads deliver for exactly as long as the card is charged. Paid social resets to zero the moment you pause. Marketplace visibility is priced by the marketplace and re-priced whenever they like.

Organic is the only one where work you paid for eighteen months ago still produces enquiries this month at no marginal cost. That doesn't make it free — our SEO starts at ₹75,000/mo, and smaller sites at ₹40,000/mo, plus real time from your team. The difference is that the asset outlives the invoice.

That gap widens over time, and in the wrong direction for paid: auction prices go one way. Every year there are more bidders on the same commercial queries, and your cost per lead climbs unless you do something about it. A page that ranks costs the same whether it gets 200 clicks a month or 2,000.

It also caps the damage from a bad quarter. If your entire pipeline is paid and you have to cut spend, your pipeline is a month of nothing. Organic keeps producing while you fix the budget.

Placement is still the cheapest trust you'll buy

Buyers notice where you appeared. A result that Google chose to rank, or that an AI answer chose to cite, carries a different weight from one labelled Sponsored — not because organic is morally superior, but because someone else decided to show it.

There's a second-order effect that shows up in every account eventually. Brand queries convert better than generic ones, because the person searching has already made a decision. What SEO does is create those people: someone reads your comparison page in March and searches your name in July. Attribution will credit the brand search and quietly ignore the page that caused it, which is why organic tends to be underrated by exactly the teams measuring it most carefully.

And being the source an assistant names is now a distribution channel in itself. It doesn't send much traffic. It sends a small number of people who arrive already told you're the answer.

Where SEO genuinely matters less now

This is the section the rest of the internet leaves out. SEO isn't uniformly important, and the AI summary layer has hit some query types far harder than others.

• If more than half your organic traffic sits in the first two rows, your next eighteen months will be unpleasant and no agency can fix that with more articles.
• The test is simple: pull your top 20 organic landing pages and mark which ones have ever produced an enquiry. That list, not the traffic chart, tells you how important SEO actually is to your business.
• If almost none of them have, the answer isn't necessarily *stop*. It's usually *you've been ranking for the wrong half of your market* — see where SEO fits in the wider plan.