The thread has 347 comments and two kinds of people in it. One group posts a Search Console screenshot with a cliff in it and says the game is over. The other group says their agency client base has never been healthier and the panic is content marketers discovering that free traffic was always on loan. Both are telling the truth. That is the part nobody in the thread works out.

So, the question as asked. Is SEO dead? No. Search volume is not falling. What is happening is a redistribution: clicks are being pulled away from informational queries and concentrated into commercial, branded, local and transactional ones, and whether that is a catastrophe or a non-event for you depends almost entirely on which of those your traffic is made of. That is not a diplomatic middle position. It is a measurable property of your own Search Console export, and you can have the number by the end of the afternoon.

What follows is the evidence, the places where Google and the measurement firms flatly contradict each other, a method for scoring your own exposure, and the arithmetic to use if the real question is “should I keep paying for this”.

The part of SEO that is actually dead

The headline number comes from SparkToro’s analysis of Similarweb panel data: 68.01% of US Google searches ended without a click to the open web across January to April 2026, against 60.45% in 2024. Push the series back and you get roughly 49% in 2019 and around 45% in 2016. The honest version of that claim is narrower than the version that gets quoted: those are different panels from different vendors in different years, which the analysis says outright, so the absolute figures are not strictly comparable. The direction has been consistent across every panel anyone has run for a decade.

Underneath the aggregate, two studies isolate the mechanism. Pew Research Center instrumented the browsing of 900 US adults and captured 68,879 real Google searches in March 2025: where an AI summary appeared, users clicked a traditional result on 8% of visits, against 15% where none appeared. Ahrefs, comparing 300,000 keywords before and after the AI Overview rollout, attributed a 34.5% CTR reduction to the feature itself after controlling for the general decline in informational CTR.

The second measurable decline is publisher referral traffic, a different thing that keeps getting merged into the first. News and reference publishers monetise attention on informational queries, exactly the segment being intercepted, which is why their numbers look apocalyptic while a commercial roofing site looks fine. If your model is “rank for a definition, show ads against it”, the last two years have been an extinction event. If it is “rank for a service someone is buying”, mostly not.

One more figure, because it gets the blame wrongly assigned: AI Mode, Google’s fully conversational surface, was 0.34% of searches in that same 2026 window. A rounding error. AI Overviews, sitting in ordinary results pages, are where the damage is being done.

Where Google and everyone else disagree, and who I believe

In August 2025 Liz Reid, who runs Google Search, published a post stating that total organic click volume from Google to websites had been “relatively stable year-over-year” and that Google was sending “slightly more quality clicks” than a year earlier, a quality click being one where the user does not quickly bounce back. She dismissed third-party reports of collapse as flawed methodology.

Google published no underlying data with that claim. None. Not a chart, not a percentage, not a definition of the denominator. Meanwhile the studies on the other side have stated sample sizes, date ranges and methodology limitations you can argue with.

And yet I think the statement is probably literally true and completely useless to you. Total click volume is an aggregate across every query on earth, and query volume is growing. If searches rise while the click rate per search falls, total clicks sit flat while the composition underneath rearranges completely. Google is describing the mean of a distribution that is pulling apart at both ends. You do not operate at the mean. You operate on maybe four hundred queries, and the only question is whether those four hundred sit on the growing side or the shrinking side.

Also worth naming: “quality clicks” is a metric Google invented, defined and measures privately, with no independent verification possible. Treat it as a position, not a finding.

What is holding up or growing

The strongest single clue in any of the published research is a detail in the Ahrefs data: 99.2% of keywords that trigger an AI Overview carry informational intent. The feature is aimed at questions, not purchases. Which gives you a rough map of insulation:

• Transactional queries. “Book”, “buy”, “quote”, “hire”, “pricing”, “near me open now”. The user wants to complete something on a specific site. A summary is an obstacle, not a service.
• Branded queries. Someone typing your company name has already decided where they are going. Your highest-converting segment, and essentially untouched, which is why brand-building quietly became an SEO tactic rather than a separate discipline.
• Local. The map pack was eating clicks long before AI Overviews existed, so local has already absorbed its disruption. The three-pack, correct hours and live reviews still drive calls that show up in a phone log.
• Commercial investigation. “Best”, “vs”, “alternative to”, “reviews”. This is the contested middle. Summaries do appear, but the user is comparison shopping and usually wants to see the actual thing, at the actual price, with the actual reviews. Clicks here have softened, not collapsed.
• Deep long-tail and situational queries. Anything specific enough that a generic answer is unsatisfying: “can I claim R&D relief on a failed prototype”, “does this part fit a 2019 model”. Volume per query is tiny, intent is enormous.

None of that is a promise. AI Mode handles comparison and shopping intent, and Google has every commercial reason to push it there. But today the measured interception sits overwhelmingly on the informational end, and any strategy argument that does not separate the two is people shouting past each other.

The new channel: tiny, and unusually good

Referral traffic from AI assistants is real and small. The SparkToro analysis puts AI tools collectively at under 1% of all traffic sent to websites. Anyone telling you to reallocate a serious budget share to chase it on volume grounds is selling something.

The interesting part is quality. Adobe Analytics, working from over a trillion visits to US retail sites, reported that across the 2025 holiday season AI-sourced traffic converted 31% better than other traffic sources, with visitors 33% less likely to bounce, spending 45% longer on site and viewing 13% more pages. Traffic volume from AI sources to retail was up 693% year over year, from a very small base.

You will also see a widely repeated claim that AI search visitors are 4.4 times as valuable as organic ones. That figure is Semrush’s, published June 2025, and I would not take it into a board meeting: it comes from a vendor selling AI visibility tooling, the published methodology is thin, and the piece it sits in is substantially a set of projections rather than observations. The Adobe numbers are better sourced and point the same direction, so use those.

The explanation is boring. The assistant already did the filtering: by the time someone clicks through they have described their situation in a paragraph, been given three candidates and picked one. Warmer than a position-four blue link. It also caps the volume by design, because the assistant sends one or two people where a results page sent ten.

How to diagnose your own exposure in an afternoon

Stop reading industry averages about your business. Here is the actual procedure.

1. Export 16 months of query data from Search Console. Sixteen, not twelve, so you can compare the same four months across two years and dodge seasonality.

2. Tag every query into one bucket with formulas, not by hand. Branded if it contains your brand or product names. Local on a city, region or “near me”. Transactional on buy, price, pricing, cost, quote, hire, book, demo, for sale. Commercial investigation on best, top, vs, versus, alternative, review. Informational on what, how, why, guide, examples, meaning, tips. The remainder goes in an “other” pile you eyeball. Twenty minutes of regex, and an imperfect split beats no split.

3. Compute clicks, impressions and CTR per bucket, year over year. Now the diagnosis, which hinges on one distinction almost nobody makes:

• Impressions flat or up, clicks down: interception. You still rank. The results page is answering instead of forwarding. This is the AI-search problem everyone is arguing about.
• Impressions down, clicks down: a ranking or demand problem. Different disease, different treatment, and it has nothing to do with AI. Check for a core update date, a competitor, a migration, or a market that shrank.
• Clicks down, conversions flat: you lost traffic that was never going to buy. Annoying for the traffic chart. Irrelevant to the business.

4. Do the same at page level. Filter for URLs that lost more than 30% of clicks while average position held within a place or two. That list is your intercepted inventory, named specifically. In most B2B libraries it is the “what is” and “how to” posts written in 2021 to build topical authority.

5. Weight it by money. Attach a value per bucket; for most service businesses a branded or transactional click is worth ten to fifty times an informational one. Recompute the loss in revenue terms. I have watched a 38% traffic decline turn into a 4% pipeline decline under this treatment, and seen a 9% traffic decline hide the loss of the two comparison pages that fed half the demo requests.

The threshold I use: if more than 60% of your organic clicks are informational and do not lead anywhere measurable, you own a genuinely shrinking asset and should be planning accordingly. If it is under 25%, the death-of-SEO discourse is largely about somebody else’s business model.

The arithmetic: SEO against paid over three years

If the real question is whether to keep funding this, do the comparison properly. Two numbers make it concrete, both from WordStream’s 2026 benchmarks, which analysed 13,474 US search campaigns across 23 industries between April 2025 and March 2026: average cost per click $5.42 and average cost per lead $66.69. Business services sat at $5.87 and $93.69; home improvement at $8.33 and $90.92; legal at $9.87 and $131.63.

First calculation. Take your commercial-intent organic clicks per month and multiply by your category CPC. A business services firm getting 1,400 such clicks is receiving roughly $8,200 a month of traffic it would otherwise rent, near $99,000 a year. Not profit, and not by itself a reason to keep the retainer, but it is the correct denominator and it usually ends the “cancel everything” conversation.

Second calculation, the honest one. Put $4,500 a month into each channel for 36 months.

Paid delivers about 48 leads a month at that CPL, from month one, flat. Over 36 months: roughly 1,730 leads for $162,000.

SEO delivers close to nothing for six months, then ramps. On a realistic curve for a competent programme in a mid-competition category, say 6 leads a month through month six, 22 a month through month twelve, 45 a month in year two and 62 in year three, you get about 1,450 leads for the same $162,000.

Read that result properly, because it goes against the usual pitch. On these assumptions SEO does not beat paid on cumulative volume inside three years. It beats it on marginal cost around month 30, when the run rate falls under $73 a lead against paid’s $93.69, and it overtakes on cumulative leads somewhere around month 48. Paid stops the day you stop paying. SEO decays over quarters rather than hours, and the asset has a resale value in a way an ad account never does.

Two things break the model, and both are worth testing. If your ramp assumption leans on informational content, discount it hard, because that is exactly the traffic the CTR studies say is being taken. If paid CPCs in your category are rising faster than inflation, the paid line degrades while the SEO line improves. Run it with your own numbers. Twenty minutes, and more useful than any thread on this subject.

What I would actually do

Keep investing, reweight, and be specific. Move budget out of top-of-funnel explainers and into the transactional, comparison, pricing and local pages that still earn clicks. Fund brand, because branded search is the one segment nothing has touched and it is now effectively an SEO input. Set a baseline this quarter for branded versus non-branded clicks and for direct traffic from new users, so next year’s argument has evidence in it. And stop reporting total organic sessions as the headline number. It stopped describing your business around 2024.

The people saying SEO is dead are usually people whose traffic was informational and unmonetised. The people saying it is fine are usually selling to buyers with wallets out. Find out which one you are before you decide.

Questions that come up next

Our clicks dropped but leads did not. Is anything wrong? Probably not, and say so loudly before someone cuts a budget over a traffic chart. Check the pages holding the leads are stable, then move session volume off the top row of the dashboard.

Should we shift the whole budget into ads? Only if you need leads inside 90 days and can live with the economics never improving. The sane version is a blend, sized so paid covers the near term while organic builds on the query types that still convert. Cancelling SEO outright is a decision to rent every future click at a price someone else sets.

How long until commercial queries get hit too? Nobody knows, including Google, and anyone giving you a date is guessing. Watch the monthly share of your own commercial queries that show an AI Overview. If that percentage starts climbing, you will see it months before it hits the click data.

Is it worth optimising for ChatGPT and Perplexity referrals? Worth doing, not worth reorganising around at under 1% of web traffic. Most of the work overlaps with normal technical and editorial SEO, so you are already paying for it.

If you want the exposure audit run rather than described, that is a service we provide. AB7 Solutions will take your Search Console and analytics exports, segment your queries by intent, separate interception from ranking loss page by page, weight the result against actual pipeline, and hand back the SEO-versus-paid model with your own CPCs in it, so the budget conversation stops being a debate about headlines. We also do the follow-on work: SEO, AEO, GEO and AIO programmes, plus the web and analytics builds behind them. Call +1 321 341 7733 or email ab@ab7solutions.com or director@ab7solutions.com. More at www.ab7solutions.com. If the audit says your exposure is low and your problem is somewhere else entirely, we will tell you that instead.

Sources: SparkToro, “In 2026, Less than One Third of Google Searches Still Send a Click” (Similarweb US panel, January to April 2026, with stated year-to-year comparability caveats); Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (900 US adults, 68,879 searches, March 2025); Ahrefs, “AI Overviews Reduce Clicks by 34.5%” (300,000 keywords, March 2024 v March 2025); Liz Reid, Google, “AI in Search: driving more queries and higher quality clicks” (6 August 2025, no supporting data published); Adobe Analytics, “AI-driven traffic surges across industries” (over 1 trillion visits to US retail sites, 2025 holiday season); Semrush AI search study, June 2025 (vendor research, projection-based, cited here with caveats); WordStream, 2026 Google Ads Benchmarks (13,474 US search campaigns, April 2025 to March 2026).