Is SEO Dead Now With AI? An Honest Answer
Search volume for SEO-service keywords fell 45-86% year on year in our data. Buying-intent search still converts. An honest answer for store owners.
Search volume for informational keywords in the SEO services niche is down between 45% and 86% year on year. That is how many people run the search, measured against last year — not a claim about where their clicks went. That is what we measured across the money keywords in our own market — the terms people type when they want an SEO service. Some pages lost nearly nine visits in ten.
So the question is fair. If you sell things online, is search engine optimisation still worth paying for, or has AI killed it?
The honest answer is in two halves, and both halves matter.
Yes, AI Overviews really are taking traffic
An AI Overview is the block of generated text at the top of a Google results page. It answers the question directly. Below it sit the ten blue links that used to get the click.
When the answer sits above the links, fewer people scroll. That is not a theory. It is the mechanism, and the drop in our data matches it.
The pages that get hit hardest are the ones that answer a single factual question. "What is a canonical tag." "How often should I blog." "What size should product images be." These posts existed to catch a question and pass the reader along. An AI Overview does the passing along without the visit.
If your traffic came mostly from posts like that, your traffic is falling and it will keep falling. Anyone telling you otherwise is selling you last year's plan.
Two more things worth saying plainly. First, the drop is uneven — it depends entirely on which queries you rank for, so a headline percentage from any agency, including this one, is a signal and not a forecast for your store. Second, some of the traffic you lost was never worth much. A reader who wanted a definition was never going to buy your kettlebells.
No, commercial search has not collapsed
Here is the other half.
People still search when they are ready to spend money, and those searches still produce clicks. In our own niche, the head terms are competitive precisely because they are still worth having. "Ecommerce SEO services" gets 1,600 searches a month in the US with a cost per click of $36. "Shopify SEO services" gets 1,300 a month at $54 per click. "Ecommerce SEO consultant" is smaller at 390 a month, but the cost per click is $112.
Advertisers do not pay $112 for a click that goes nowhere. Those prices are the market telling you what a buying-intent search is worth.
One term in our set is growing rather than shrinking: "affordable seo services", up 48% year on year at 1,300 searches a month. People are not searching less about this. They are searching with more scepticism about price.
The pattern holds outside our niche too. Someone typing "merino base layer women's" or "replacement filter for [model number]" is not looking for a paragraph of explanation. They want a product page, a price, and a delivery date. Generated summaries are bad at that job, and Google knows it, which is why those results still look like results.
The second surface: being named by an assistant
Something new has appeared alongside search, and it is easy to miss because it does not show up in your analytics.
People now ask ChatGPT, Claude, Gemini and Perplexity what to buy. "Best waterproof dog bed for a large breed." "Which US brands make office chairs under $400." The assistant answers with a short list of names.
Being on that list is a distribution channel. Not being on it is invisibility of a kind that no rank tracker will report to you, because there is no rank.
What gets a brand named is unglamorous. The assistant needs to find clear, structured information about your products, from sources it trusts, that agree with each other. That means your own site is legible, and it means other sites — reviews, roundups, forums, comparison pages, retailer listings — say the same things about you.
This is not a separate discipline with a new acronym. It is mostly the old work, aimed at a new reader.
What actually changes in how the work is done
Four practical shifts.
Fewer thin informational posts. The "what is X" article aimed at a question an AI Overview now answers is dead weight. Writing more of them is money on fire.
More commercial and product-adjacent pages. Collection pages, comparison pages, buying guides tied to real products, sizing and compatibility pages, and pages that answer the specific objection that stops a purchase. These sit close to the money and they still get clicked.
Structured data, properly. Structured data is code in the page that labels what things are — this is a product, this is its price, this is the review score, this is whether it is in stock. Both Google and the assistants read it. Most Shopify and WooCommerce stores have some of it by default and have it partly broken: prices that do not match, missing stock status, review markup that never validated. Fixing it is a technical job with a checkable result.
Being citable rather than just being ranked. Clear claims. Specifics over adjectives. Numbers with a source attached. Pages that state a fact in a sentence an assistant can lift without garbling it. Vague marketing copy cannot be cited because there is nothing in it to cite.
None of this is exotic. It is a change of target, not a change of trade.
What ChatGPT alone cannot do for your store
You can absolutely write with an LLM. We do. Being cagey about that would be silly, and the audience for this article can spot generated filler as well as we can.
But a chat model on its own cannot run your SEO, for four specific reasons.
It cannot tell you search volume. Ask it how many people search for your product term each month and it will produce a plausible number. That number is not measured. It comes from patterns in text, not from a keyword database. Every figure in this article came from a keyword data provider, not from a model.
It cannot tell you difficulty. Whether you can realistically rank depends on who currently holds those positions and how strong they are. That requires looking at the live results page and the link profiles behind it. A model has not looked.
It cannot tell you who holds the position now. Results change weekly. A model's knowledge is a snapshot, and it will not tell you when the snapshot was taken.
It cannot tell you what is broken on your store. Your duplicate collection URLs, your unindexed variants, your product schema failing validation, your 400ms server response on mobile — none of that is in any training set. It needs a crawler pointed at your domain.
The model is a good writer and a poor analyst of your specific situation. Use it as the first and not the second.
So: is it worth it in 2026?
If "SEO" to you means publishing informational blog posts and hoping, no. That was already weakening and AI Overviews finished it.
If it means being findable and credible at the moment someone decides to buy — on Google and inside an assistant — then yes, and the cost per click numbers above are the market agreeing.
The reasonable position is neither panic nor denial. Look at what your own traffic is made of. If it is mostly informational, expect it to keep falling and stop investing there. If you have commercial pages that are technically sound and genuinely useful, they are still working, and the second surface is a reason to make them clearer rather than a reason to start over.
Is SEO dead now with AI?
No, but one part of it is dying. Demand for single-question informational keywords is shrinking — we measured search-volume drops of 45% to 86% year on year across the SEO services market. AI Overviews are the most likely cause, but volume is what we measured, not attribution. Commercial and buying-intent search still converts, and cost per click on those terms stays high, which is advertisers voting with money.
Can ChatGPT do SEO?
It can do the writing. It cannot measure search volume, judge ranking difficulty, see who holds a position today, or find what is broken on your store — those need keyword data, live results pages, and a crawler. Treat it as a drafting tool with a human deciding what to draft.
Is SEO still worth it for a small ecommerce store in 2026?
It is worth it if you aim at commercial intent rather than blog traffic. The practical test: list the pages that brought you sales in the last 90 days, not the ones that brought visits. If those pages are product, collection or comparison pages, that is where the work belongs — plus clean structured data so assistants can read and name your products.
If you want an outside read on which of your pages are still earning and which are just costing you, that is exactly what a keyword and technical audit is for. Ours comes with the monthly service, and you can see what that costs before you talk to anyone.