Is SEO dead because of AI?

If you run a firm doing real revenue and your organic traffic has gone soft while your rankings look about the same, you’re probably wondering whether SEO is dying and whether the money you put into it was wasted. Here’s the short answer. SEO still works, but a page-one ranking earns fewer clicks now that Google puts an AI summary on top of a lot of searches, and you can work on that once you know where the clicks went.

Is SEO actually dead because of AI?

No, SEO is not dead because of AI, but what a ranking pays you has changed. Google’s own documentation says the best practices for SEO remain relevant for its AI features, and a page has to be indexed and eligible to show with a snippet before Google’s AI can link to it at all.

It used to be you ranked a page and the buyer clicked it and read your pitch, but now there’s an answer sitting on top of the results and a lot of buyers read it and stop. The work underneath is mostly the same work, but the goal has shifted from getting the click to getting quoted, and the click comes after that if it comes at all.

Google said as much in its guide to optimizing for generative AI features, last updated July 2026, where it calls this work “still SEO” and tells site owners they don’t need AI text files, special schema or content chopped into tiny pieces to show up. I read that as Google telling you the fundamentals still decide who gets in.

I’d rather you hear it from Google than from me, cause I sell this work. My honest opinion is the loudest “SEO is dead” voices tend to be selling whatever comes next, and the people saying nothing changed tend to be selling the same retainer they sold you years ago.

How much traffic are AI Overviews taking from websites?

A lot, on the searches where they show up. A Pew Research Center study of 68,879 real Google searches found people clicked a regular search result on 8% of visits when an AI summary appeared, against 15% when it didn’t, and only 1% of those visits included a click on a link inside the summary itself.

Pew tracked the actual browsing of 900 U.S. adults in March 2025, so these are clicks people really made and not answers they gave on a survey. People were also more likely to end their browsing session entirely after a search with an AI summary, 26% of the time against 16% without one, so for a good chunk of people the summary is where the search ends.

Which searches lose the most?

Question searches lose the most, and that’s bad news for anyone who sells a considered service. Researchers at Washington University in St. Louis ran 55,393 Google searches between March and April 2026 and found AI Overviews on 13.7% of all queries but on nearly 65% of question-form queries, with an average overview citing about eight sources.

“Who handles truck accident cases in Houston” is a question and so is “how much does a custom home cost per square foot,” so a firm whose traffic came from answering questions like those is gonna feel this first.

Pew’s data is from early 2025 and habits keep moving, so I’d treat the exact percentages as a snapshot and watch which way they point.

Do AI tools still pull their answers from search rankings?

Mostly yes, they do. A University of Toronto study of ChatGPT found that every one of the 22,315 web pages it cited had come back in its own searches first, so if a page doesn’t show up when the AI goes searching it doesn’t get quoted, and ranking is still how you get in the door.

The same paper, by Mahe Chen and Nick Koudas, found ChatGPT cited 43.8% of the pages it ranked first in its own search results and only 3.5% of the pages it ranked 11 to 20. That’s ChatGPT’s own ranking and not Google’s, but the top spots still soak up most of the attention.

Google’s AI Overviews follow a looser version of that. A TripleDart study of 11,002 pages across 1,000 B2B software keywords found pages in position 1 were cited 55% of the time and pages in position 9 were cited 34% of the time. Ranking higher still helps, but a ninth-place ranking never used to be worth much of anything.

Then why do so many AI Overview citations come from outside the top 10?

Because Google’s AI doesn’t only look at the results for your exact search. Ahrefs looked at 863,000 keywords and 4 million AI Overview links and found that only 38% of cited pages also ranked in the top 10 for the same query, down from 76% in its July 2025 study, though Ahrefs changed how it detects citations between the two so the drop isn’t a clean comparison.

Google’s guide explains part of it. It describes query fan-out, where the model runs a set of related searches on its own to gather more information, which means the AI might quote you for a side question you rank well on even when you’re nowhere for the main one. That’s why I think covering a whole topic now beats owning one keyword, and it’s the reason I build sites around topic clusters and pillar pages instead of one page per keyword.

What still works on Google?

The fundamentals still work on Google, and Google keeps saying so. Indexed, crawlable, clearly written, specific and current pages are what its AI features can find and quote, and what’s different is building the page so a machine can lift a clean answer out of it.

Where should the answer go on a page?

Up top, in plain words. The AI needs a passage it can quote and a buyer skimming on a phone needs the same thing, and TripleDart found cited pages covered more of the search question across their H2 headings than skipped pages did, 0.62 against 0.59. That gap is small, I’ll admit, but it lines up with what Google keeps telling people, which is to write for the person asking and make the answer easy to find.

Every important page should open with the direct answer and then back it up with the stuff only you have, like your process, your terms, your credentials and real numbers you can stand behind. For the longer version of how Google’s AI picks what it quotes, read my post on how to rank in Google AI Overviews.

How old is too old for a page?

Older than a year starts to hurt, at least on ChatGPT. An AirOps study of more than 4,000 pages ChatGPT cited, across 900 high-intent queries in 15 industries, found that pages that hadn’t been updated in over a year were more than twice as unlikely to be cited. That study looked at ChatGPT and not Google, but a service page you wrote in 2021 and never touched is about the easiest page on the internet to replace.

When I say update I mean the facts, the prices, the process and the examples, and then a short note saying what changed and when, cause swapping the date at the top and leaving everything else alone is just a fib to the reader.

Does getting cited still bring clicks?

It looks like it does. Seer Interactive looked at more than 3,100 queries from June 2024 to September 2025 and found websites cited inside AI Overviews saw 35% higher organic click-through rates and 91% higher paid click-through rates than sites that weren’t cited, and that’s the part people skip right past when they say SEO is dead.

Now, that’s a correlation and not proof the citation caused the clicks, cause brands that get cited tend to be stronger brands to begin with. I still think it tells you where the clicks that are left are going, and they’re going to the names in the answer.

What still works on Bing, and why does it matter now?

Bing matters now because it feeds AI answers you can’t see in Google’s tools. Bing Webmaster Tools added an AI Performance report in February 2026 that shows how often your pages are cited in Microsoft Copilot, AI-generated summaries in Bing and select partner integrations, plus a sample of the searches the AI ran to find you.

If your firm never opened Bing Webmaster Tools, that was fair back when Bing was a small slice of your traffic, but I think it’s a bad call now, cause this report counts your AI citations page by page and Search Console doesn’t break AI citations out that way.

Bing’s announcement says you get total citations, the average number of your pages cited each day, a sample of grounding queries (the phrases the AI used when it went looking), citation counts per page and a timeline, and I’d call that the closest thing to a free AI visibility scoreboard you’ll find right now.

Does Bing ranking affect ChatGPT?

Probably some, though the evidence is thin. A Seer Interactive study from February 2025 matched more than 500 SearchGPT citations against search results for about 100 questions and found 87% of the citations matched Bing’s top organic results, against 56% for Google. Seer itself calls the sample limited and the finding directional, and ChatGPT’s search has changed since then.

I won’t tell you Bing is the secret to ChatGPT, cause one small study can’t prove that. What I will tell you is that Bing Webmaster Tools is free, the report gives you real numbers, and if you skip it you could be leaving a door open for a competitor to walk through first.

What stopped working in SEO?

The volume game stopped working. Publishing dozens of near-identical pages to catch every keyword variation, writing to a word count and stuffing keywords into thin copy now range from useless to risky, and Google’s spam policies name several of those habits directly.

Google defines doorway abuse as pages “created to rank for specific, similar search queries.” It defines scaled content abuse as many pages generated “for the primary purpose of manipulating search rankings,” no matter how they were made, and its AI guide goes a step further and warns that making separate pages for query variations mainly to manipulate rankings or its AI answers breaks that same scaled content policy.

Which old habits should you drop?

A few, while you’re at it. Google says there’s no ideal page length, so a 3,000-word page isn’t better for being 3,000 words, and it says its AI systems understand synonyms, so you don’t need a page for every long-tail variation either. It also says Google Search doesn’t use AI text files like llms.txt, so if somebody sold you one as an AI ranking trick I wouldn’t expect it to move anything on Google.

Generic copy belongs on that list too. If your site reads like every other site in your industry, a model has no reason to pick you over the next firm and a buyer doesn’t either, and I wrote a whole post on what helpful content means now cause I think it’s the part that gets skipped the most.

How should an established company measure SEO now?

Measure clicks and citations side by side. Google Search Console shows what your organic search traffic is doing, Bing’s AI Performance report shows your AI citations by page, and a fixed list of real buyer questions, run on a schedule through the AI tools, shows whether your name is in the answers your buyers read.

That question list is the piece I’d start with if you don’t have one. Write down 50 questions your buyers actually ask before they hire someone like you, typed the way they’d type them, then run them through ChatGPT, Claude, Perplexity, Gemini and Google’s AI and write down who gets named. Run it again a month later, cause AI answers change a lot day to day and one run tells you very little.

For the bigger picture on the AI side, my post on what AI visibility is and how it’s measured is the place to start, and how to get your business cited by ChatGPT has the steps for that one tool.

AI answers send fewer visitors, so each one has to count for more, and that’s why I also wrote about conversion rate optimization for the traffic you still get.

What does it cost to wait?

Waiting usually costs you slowly and not all at once. Pew found clicks on regular results roughly halve when an AI summary shows up, the Washington University team found those summaries on nearly 65% of question searches, and AirOps found pages left alone for a year were more than twice as unlikely to be cited by ChatGPT. None of that shows up as one bad month. What you see is a traffic line that sags a little every quarter and leads that feel a bit softer while nobody can say why.

That slow leak is the common outcome. The rarer and worse one is a competitor becoming the name the AI gives for your service in your city, and I can’t tell you how likely that is for your firm without looking.

These answers aren’t locked, though, they move day to day, so a firm that fixes its pages this quarter has a real shot at answers it’s missing today. This week, type your ten most common buyer questions into ChatGPT and Google and write down who gets named, cause that list is your starting point.

How I know this works

I run my own ecommerce brand and I did this work on it first. In July 2026 it ranked #1 on Google for its main “best” product search and it was the first product Google’s AI named for that search when I checked, and that same month 18 of its 22 target terms were in the top 3 on Google.

Google Search Console showed it on page one for more than 400 terms, on a three-month average to June 2026. Bing’s AI Performance report shows its AI citations went from 9 a month in April 2026 to 1,275 a month in August 2026. Those are my own numbers from my own accounts, pulled straight from those tools for those dates.

It’s one brand, in ecommerce, and it runs paid ads too, so not all of its growth is organic and your results will be different. You can see the numbers behind my earlier results on my homepage, along with how each one was measured.

What are your options if your organic traffic is slipping?

You’ve got four realistic options, and which one’s right depends on your budget, your team and how many pages you’ve got. Here’s how I’d compare them, including the one I sell.

If you’re a smaller firm with a handful of pages, I think in-house is probably right, and the rest of this blog is written to help you do it.