Is SEO Dead? Search, AI Overviews, and Zero-Click in 2025-2026

What the click data actually shows, and which viral SEO stats do not survive their own source

Research date: August 9, 2026
Scope: Google web search in the US and global aggregate; English-language B2B relevance. Studies limited to one country, device, or vertical are labeled as such and never generalized into a universal rule.
Evidence standard: first-party disclosures from Google (Search Central documentation, Search Liaison statements) and Alphabet (quarterly earnings) first; transparent large-sample clickstream and CTR datasets with a stated method second (Pew Research, SparkToro/Datos/Similarweb, Ahrefs); vendor studies with partial methods third; content-farm statistics with no auditable source are not accepted as proof.

Executive summary

SEO is not dead, and search is not being killed by ChatGPT. Both stories are laundered. The honest version is more specific and more useful: the number of searches that send a click to the open web is genuinely and structurally falling, but Google itself is bigger and more profitable than ever, and most of the viral "ranking factor" stats that people optimize against are either a decade old, correlation dressed as causation, or contradicted by Google's own words.

What holds up:

• Zero-click is real and it is now the majority, by a wide margin. SparkToro's clickstream analysis found 68.01% of Google searches ended without a click to the open web in the first four months of 2026, up from about 45% a decade ago. That is a transparent, method-disclosed finding, and it is the single most important number in the category. (SparkToro, June 2026)
• AI summaries are associated with fewer clicks, measured by a neutral panel, not just by vendors. Pew Research, using the browsing data of 900 US adults across 68,879 searches in March 2025, found users who saw an AI summary clicked a traditional search result in 8% of visits, versus 15% when no summary appeared. Users clicked a link inside the summary itself in just 1% of visits. This is the cleanest available measurement because Pew sells nothing. (Pew Research, July 22, 2025)
• "Google is dying" is contradicted by Google's own filings. Alphabet reported Q2 2026 "Google Search and other" revenue of 63.3 billion US dollars, up 17% year over year, inside the AI Overviews era. Google held roughly 91% of the global search-engine market in July 2026. A business losing to ChatGPT does not grow its core line 17%. (Alphabet Q2 2026 earnings, July 22, 2026; StatCounter, July 2026)
• The famous "#1 result gets ~28% of clicks" figure is real, but it is a 2022 study and an average that hides enormous variance. Backlinko's analysis of about 4 million results found the #1 organic result at 27.6% CTR and only 0.63% of searchers clicking to page two. It is a legitimate large sample. It is also four years old, predates AI Overviews entirely, and averages across query types that behave nothing alike. (Backlinko, 2022)
• "Backlinks don't matter" and "E-E-A-T is a ranking factor" are both wrong, in opposite directions, and both are settled by Google's own statements. Google's Gary Illyes said links are not a top-three ranking factor and "haven't been for some time," while never saying they stopped mattering. Google's documentation states plainly that "E-E-A-T itself isn't a specific ranking factor." Repeating either as its opposite is the error. (Search Engine Land on Illyes, Sept 2023; Google Search Central)
• "Longer content ranks better" is correlation wearing a causation costume. Google's John Mueller has said repeatedly that word count "is not a ranking factor." Longer pages sometimes correlate with better rankings because they more often satisfy the query, not because length is rewarded. (Search Engine Journal on Mueller)

The practical conclusion for a founder deciding whether to still invest here: search is not dead, but its economics changed. You are now competing for a shrinking pool of clicks per search, so the return has shifted away from "rank for a keyword and harvest traffic" toward being the source that gets cited or that answers the high-intent, decision-stage query a summary cannot fully resolve. Stop optimizing against 2022 CTR curves and decade-old "ranking factor" lists. The channel still reaches nearly everyone who searches. It just sends fewer of them to your site, and rewards being the answer rather than the tenth blue link.

Claim-by-claim verification

Provenance flags

• The "~28% CTR for #1" figure is a 2022 average, not a 2026 fact. Backlinko's study is a real ~4M-result analysis, but it is four years old, predates AI Overviews, and averages across informational, navigational, and transactional queries that have wildly different curves. Anyone quoting it as today's CTR is quoting a pre-AI-Overviews number. Advanced Web Ranking updates a live curve monthly; use a dated live source, not a static 27.6%. (Backlinko, 2022)
• The "53% of traffic is organic" figure is a single 2019 BrightEdge study. It is repeated across hundreds of downstream posts as a standing 2026 fact. It is not. It is one vendor's 2019 clickstream, direct traffic removed, and it varies by more than 20 points across verticals. Cite it as 2019 context or not at all. (BrightEdge, 2019)
• "Google says AI Overviews send higher-quality clicks" is an unsupported first-party counter-claim. Google has publicly asserted that AI Overviews increase or improve clicks, but has released no dataset to check it against, while independent measurement (Pew, Ahrefs) points the other way. Do not repeat Google's claim as measured; note it as an unverified company statement. (Ahrefs, 2025)

What the primary sources actually say

Zero-click: the one number that is not in dispute

SparkToro has tracked this for a decade, and its 2026 update is the load-bearing finding of this report. Using a Similarweb cross-device panel, it found 68.01% of Google searches ended without a click to the open web in January through April 2026. It puts the decade-ago figure near 45%, a 23-point rise. The historic comparison mixes panels (Jumpshot for 2016-2019, Datos for 2024, Similarweb for 2026), so the trend line is directional rather than a clean apples-to-apples series, and SparkToro says so plainly. The direction is not ambiguous: the share of searches that send a click out is falling, and it crossed the halfway mark years ago. (SparkToro, June 2026)

The earlier viral "~60% of searches end without a click" belongs to SparkToro's 2024 study, built on Datos (now a Semrush company) clickstream. That study found 58.5% of US searches were zero-click and that for every 1,000 US searches, only 360 clicks reached the open web, with nearly 30% of clicks going to Google-owned properties. So the "60%" number was real, US-specific, and a 2024 figure. The 2026 successor is 68%. (SparkToro, 2024)

One correction the 2026 study makes to its own prior work: paid clicks were undercounted in 2024 because an unusually high share of the Datos panel ran ad blockers. Treat the exact paid-click split with that caveat. The zero-click headline is not affected.

AI Overviews and clicks: the neutral measurement points down

The most credible source here is not a search vendor. Pew Research analyzed the actual browsing data of 900 US adults across 68,879 searches in March 2025. When an AI summary appeared, users clicked a traditional search result in 8% of visits. When no summary appeared, they clicked in 15% of visits, nearly twice as often. Users clicked a link inside the AI summary itself in just 1% of visits. Users were also more likely to end their browsing session after a page with a summary (26%) than without one (16%). Pew sells no SEO product, which makes this the cleanest read available. (Pew Research, July 22, 2025)

Ahrefs, a search vendor, reached a compatible conclusion by a different route. Across 300,000 keywords it compared position-one CTR before and after the US rollout of AI Overviews. Position-one CTR for AI Overview keywords fell from 0.073 in March 2024 to 0.026 in March 2025. After forecasting what the CTR "should" have been absent AI Overviews (0.040), it attributed a roughly 34.5% reduction in position-one CTR to the presence of an AI Overview, for informational queries. This is a modeled estimate, not a direct measurement, and it is limited to informational intent, where summaries are most common. (Ahrefs, 2025)

Two honest datasets, one neutral and one vendor, pointing the same direction. What is not supported is any single portable percentage. The Pew drop (15 to 8) and the Ahrefs drop (~34.5% on position one, informational) are measuring different things on different populations. The defensible statement is "AI summaries are associated with materially fewer clicks to sites, especially on informational queries," with the sample attached, not "AI Overviews cut traffic by X%."

Note the boundary the loudest version of this claim crosses: Pew measured click behavior on the results page, not sitewide traffic loss. Panel-based studies now measure large publisher declines directly. Chartbeat, across about 2,500 publisher sites, found US organic search traffic down 38% from November 2024 to November 2025, and Press Gazette reported global publisher Google traffic down roughly a third in 2025. But attributing all of that decline to AI Overviews specifically still requires isolating the variable from ranking changes, seasonality, and platform shifts, which the public data does not do. The declines are measured; the clean attribution to one cause is not.

Google's market position: growing, not dying

The "Google is losing to ChatGPT" story does not survive Google's own filings or independent share data. Alphabet's Q2 2026 results reported "Google Search and other" revenue of 63.3 billion US dollars, up 17% year over year, with total company revenue of 119.8 billion, up 24%. Search is Alphabet's largest revenue line and it accelerated, not declined, in the middle of the AI Overviews rollout. (Alphabet Q2 2026 earnings, July 22, 2026)

On usage share, StatCounter put Google at 91.31% of the global search-engine market in July 2026, with Bing at 4.47% and Yahoo at 1.24%. In StatCounter's separate Google-vs-ChatGPT referral comparison, Google was 99.78% and ChatGPT 0.22%. StatCounter measures referral traffic to sites in its network, which is a specific method with its own limits, and it is not a measure of "where people ask questions." But as a check on the "ChatGPT is eating Google" claim, the direction is unambiguous: as a source of web referrals, chatbots are a rounding error next to Google today. (StatCounter, July 2026; StatCounter Google vs ChatGPT, July 2026)

The nuance worth holding: Google can grow revenue and share while still sending fewer clicks per search to the open web. Those are not in conflict. The threat to publishers is real, and the threat to Google's dominance is, so far, not.

AEO / GEO: a real shift, a premature obituary for SEO

The claim that answer-engine or generative-engine optimization has "replaced" SEO conflates two different things. The genuine shift is answer-in-place: Google resolving more queries on the results page, via AI Overviews, featured snippets, and knowledge panels, which is exactly what the 68% zero-click number captures. The overstated part is the migration to standalone chatbots. SparkToro's 2026 analysis notes AI tools send less than 1% of all traffic out to the open web, and Google's own AI Mode was only 0.34% of searches in January through April 2026. Google said at I/O 2026 that AI Mode had passed 1 billion monthly users with queries more than doubling each quarter, so the trajectory is steep, but as of mid-2026 the behavior change is concentrated inside Google's results, not in a mass exodus to a new engine. (SparkToro, 2026)

Vendors selling "AI visibility" tools have a commercial interest in the stronger claim. Semrush, for instance, published a study of 200,000 AI Overviews and now markets an enterprise AI-optimization suite around tracking brand presence in AI answers. The study is real; the framing that it proves SEO is obsolete is a sales position, not a finding. The measurable reality is that optimizing to be cited by an answer, and to win the queries an answer cannot fully resolve, matters more than it did in 2022, and that ranking-for-raw-traffic matters less. That is an evolution of SEO, not its replacement. (Semrush AI Overviews study)

CTR by position: a real curve, a stale number

The instinct behind "the #1 result gets ~28% and nobody goes past page one" is correct in shape and wrong in precision. Backlinko's 2022 analysis of about 4 million results found #1 at 27.6% CTR, the top three at 54.4% combined, and only 0.63% of searchers clicking anything on page two. The #1-dominates and page-two-is-a-graveyard patterns hold up and repeat across studies. (Backlinko, 2022)

But the specific 27.6% is a 2022 average from before AI Overviews existed, blended across query types that behave nothing alike. A navigational query for a brand name sends far more than 28% to the top result; an informational query now often sends far less, because a summary sits above the first organic link. Advanced Web Ranking maintains a monthly-updated CTR curve segmented by SERP feature, which is the right shape of source; its current exact position-one figure was not independently retrieved in this session and is flagged below. The rule to publish is "position one leads by a large margin and page two is negligible," dated, not "position one gets 27.6%" as a standing fact.

Backlinks and E-E-A-T: settled by Google's own words, in opposite directions

These two claims fail for the same reason: people restate Google's position as its opposite.

On links, Google's Gary Illyes said at Pubcon in September 2023 that he does not agree links are a top-three ranking factor and that "they haven't been for some time," adding that people overestimate their significance. He explicitly did not say links stopped mattering; he called them important. This is a real change from Google's 2016 messaging, when Andrey Lipattsev named links among the top three factors. So "backlinks are the number-one factor" is outdated, and "backlinks don't matter anymore" overshoots the actual statement. The defensible read: links are one signal among many, not the dominant lever they were once described as. (Search Engine Land, Sept 2023)

On E-E-A-T, Google's documentation could not be clearer: "While E-E-A-T itself isn't a specific ranking factor, using a mix of factors that can identify content with good E-E-A-T is useful." E-E-A-T (expertise, experience, authoritativeness, trustworthiness) is a concept Google's human quality raters use to evaluate whether the algorithm is producing good results. It is not a dial in the ranking system. "Improve your E-E-A-T score" describes something that does not exist as a score. Trust, Google says, is the most important of the four, and the concept matters most on health, financial, and safety topics. But it is a lens, not a factor. (Google Search Central; Google Search Central Blog, Dec 15, 2022)

Word count: correlation, not causation

Google's John Mueller has said, across multiple channels, that "the number of words on a page is not a quality factor, not a ranking factor," and on Reddit, "Word count is not a ranking factor. Save yourself the trouble." Longer content does sometimes correlate with better rankings, because a longer page more often fully answers the query and more often attracts links, both of which relate to real signals. But length itself is not rewarded. Writing to a word count to rank is optimizing the proxy instead of the thing. (Search Engine Journal on Mueller; Search Engine Roundtable)

Primary versus inferred: what can go behind your name

First-party disclosures or transparent large-sample datasets, safe to state with method attached

• Google Search "and other" revenue was 63.3 billion US dollars in Q2 2026, up 17% year over year; total Alphabet revenue 119.8 billion, up 24%. State the quarter. (Alphabet, 2026)
• Google held 91.31% of the global search-engine market in July 2026 (Bing 4.47%); in StatCounter's referral comparison Google was 99.78% vs ChatGPT 0.22%. State that StatCounter measures referral traffic. (StatCounter, July 2026)
• 68.01% of Google searches ended without an open-web click in Jan-Apr 2026 (Similarweb panel); 58.5% of US searches were zero-click in 2024 (Datos panel). State the panel and that the multi-year trend mixes panels. (SparkToro, 2026)
• When an AI summary appeared, users clicked a traditional result in 8% of visits vs 15% without; clicked a link inside the summary in 1%; ended their session 26% vs 16%. Sample: 900 US adults, 68,879 searches, March 2025. (Pew, 2025)
• Ahrefs modeled a ~34.5% reduction in position-one CTR for AI Overview (informational) keywords, March 2024 to March 2025, across 300,000 keywords. Label it a modeled estimate for informational intent. (Ahrefs, 2025)
• Google states E-E-A-T "isn't a specific ranking factor"; word count "is not a ranking factor" (Mueller); links are "not a top-three ranking factor" but remain important (Illyes, 2023). (Google Search Central; Search Engine Land)
• Backlinko (2022, ~4M results): #1 at 27.6% CTR, top three 54.4%, page two 0.63%. State the year and that it predates AI Overviews. (Backlinko, 2022)

Inference or myth: do not present as an established fact

• "SEO is dead." (Organic clicks per search are shrinking; the channel is not dead.)
• "AI Overviews cut traffic by X%" as a single portable number.
• "Google is losing to ChatGPT." (It grew Search revenue 17% and holds ~91% share.)
• "AEO/GEO replaced SEO." (Real shift is answer-in-place inside Google; chatbots are under 1% of referrals today.)
• "The #1 result gets 27.6% of clicks" as a current, universal figure.
• "Backlinks don't matter" and, in reverse, "backlinks are the top factor."
• "E-E-A-T is a direct ranking factor" or "improve your E-E-A-T score."
• "Word count is a ranking factor" / "longer content ranks better" as causation.
• "53% of traffic is organic" as a standing 2026 benchmark (it is a 2019 study).

Dated timeline

Steelman: why "SEO is basically over" might be closer to right than we allow

The strongest case for the pessimists has three honest parts, and the click data is on their side even where the "dead" headline is not.

1. The trend line only points one way. Zero-click went from ~45% to 68% in a decade and rose 10 points between the 2024 and 2026 studies alone. There is no visible mechanism by which Google reverses this, because answering in place serves Google's own incentives: more queries, more time on Google, more ad surface. A channel where the majority of uses now send no click, and that majority is still climbing, is genuinely shrinking as a traffic source even if the search box is more popular than ever.

1. AI Mode is the accelerant that has not hit yet. AI Mode was only 0.34% of searches in early 2026, but Google reported it passing 1 billion monthly users with queries doubling each quarter. If AI Mode becomes the default surface, the informational-query click, already the most suppressed, could approach zero. The current 68% could look quaint. The people planning for "SEO is over" are extrapolating a curve that Google itself is steepening.

1. The neutral measurement, not just the vendors, shows suppression. It would be easy to dismiss Ahrefs and Semrush as talking their book. But Pew sells nothing, and Pew found the click rate cut nearly in half when a summary appeared, with in-summary clicks at 1%. When the disinterested source and the interested sources agree on direction, the effect is real.

The honest synthesis:

As a channel for harvesting free clicks from informational queries, SEO is in structural decline, and the decline is accelerating. As a channel for reaching buyers at the moment of a high-intent, decision-stage query, and for being the source an answer cites, it is not dead and Google is not losing. The correct posture is neither "SEO is dead" nor "nothing changed." It is that the return has moved from ranking-for-traffic to being-the-answer, and the businesses that keep optimizing for 2022's click curve are the ones who will feel like it died.

The most authoritative sources

1. Pew Research: Google users click less when an AI summary appears - July 22, 2025; 900 US adults, 68,879 searches, March 2025; the only neutral, non-vendor measurement of AI-summary click effects. The reference dataset for this report's AI-Overview claims.
2. SparkToro: less than one third of Google searches still send a click - June 2026; Similarweb panel; 68.01% zero-click for Jan-Apr 2026, with transparent caveats about mixed panels. The load-bearing zero-click source.
3. Alphabet Q2 2026 earnings - July 22, 2026; Search revenue 63.3B, up 17%; the first-party rebuttal to "Google is dying." Links to Alphabet's official earnings release.
4. StatCounter search-engine market share - July 2026; Google 91.31%; independent share data, referral-based method.
5. Google Search Central: creating helpful content - current; the primary statement that E-E-A-T "isn't a specific ranking factor."
6. Ahrefs: AI Overviews reduce clicks by 34.5% - 2025; 300,000 keywords; a transparent, if modeled, vendor estimate consistent with Pew's direction.
7. SparkToro/Datos 2024 zero-click study - 2024; the origin of the "~60% zero-click" figure (58.5% US), with the ad-blocker caveat on paid clicks.
8. Search Engine Land on Gary Illyes: links not a top-three factor - September 2023; the primary-sourced quote behind the backlinks verdict.
9. Backlinko: 4 million search results CTR study - 2022; the standard CTR-by-position numbers, clearly pre-AI-Overviews.
10. BrightEdge: organic share of traffic increases to 53% - 2019; the origin of the "53% organic" stat, for provenance.
11. Search Engine Journal on Mueller: word count not a quality factor - the primary-sourced statement behind the word-count verdict.
12. Google Search Central Blog: E-A-T gets an extra E - December 15, 2022; confirms E-E-A-T as a rater concept.

Open questions the primary sources do not answer

• A neutral, panel-based measurement of sitewide traffic loss attributable specifically to AI Overviews, isolated from ranking and seasonality. Pew measured on-SERP click behavior, not total site traffic.
• Whether Google's claim that AI Overviews send "higher-quality" clicks is true. Google has released no dataset; independent measurement points the other way.
• A current, method-disclosed CTR-by-position curve segmented by SERP feature and query intent, replacing the 2022 Backlinko averages. Advanced Web Ranking maintains one, but its exact current figures were not independently retrieved this session.
• The real trajectory of AI Mode adoption and its click effect, given Google's "1 billion users, doubling quarterly" statement versus its 0.34%-of-searches share in early 2026.
• How much of the zero-click rise is AI Overviews specifically versus instant answers, knowledge panels, and other in-SERP features that predate them.
• A probability-sampled measure of where people now go to ask questions (Google vs chatbots), rather than referral-based proxies that only see traffic sent to the open web.

Verification pass, August 9, 2026

The five figures the first pass could not confirm were run through a second verification. One is verified at the filing level, two are strengthened, one is corrected upward, one stays fragmented. None of the report's verdicts moved; the pass hardened the click-decline evidence.

Verified:

• Alphabet Q2 2026 Search revenue of 63.3 billion US dollars, up 17%, is confirmed against Alphabet's own earnings release. "Google Search and other" grew 17% to 63.3 billion; total Alphabet revenue was 119.8 billion, up 24%. (Alphabet Q2 2026 earnings release)

Strengthened, and pointing harder in the report's direction:

• "Publishers lost N% of traffic to AI Overviews" is no longer unsourced. Multiple panel-based studies now measure it and converge on a large decline. Chartbeat, across about 2,500 publisher sites, found US organic search traffic down 38% from November 2024 to November 2025, with smaller sites losing 60% of search referrals over two years versus 22% for large publishers. Press Gazette reported global publisher Google traffic down roughly a third in 2025. Digiday cited data linking AI Overviews to a 25% drop in publisher referral traffic. The figure varies by panel and method (25 to 38% traffic, worse for small sites), and none fully isolates AI Overviews from ranking changes and seasonality, so the causation caveat stands. But the effect is measured, not asserted. (Chartbeat via DesignRush; Press Gazette; Digiday)

Corrected upward:

• Ahrefs updated its own number as AI Overviews expanded. The 34.5% position-one CTR reduction cited in the body is the 2025 figure. Ahrefs' February 2026 study (300,000 keywords plus aggregated Search Console data, December 2023 to December 2025) reports position-one CTR for AI Overview keywords falling from 7.3% to 1.6%, a roughly 58% reduction. Direction unchanged, magnitude grown, consistent with AI Overviews covering more queries over time. (Ahrefs, February 2026)

Still fragmented, use a dated live source:

• Advanced Web Ranking's single current position-one CTR figure cannot be pinned to one clean primary number. The 2026 landscape confirms the variance thesis, clean SERPs land position one near 30 to 40% and AI-Overview SERPs near 10 to 19%, but those figures circulate mostly through vendor and content-farm posts, not one auditable primary table. The CTR-by-position verdict continues to rest on Backlinko (2022) for the shape, with the instruction to read a dated, feature-segmented live curve rather than quote a static number.
• Semrush's specific 200,000-AI-Overviews sub-findings were not extracted and are used only to illustrate that AI Overviews are widespread and that vendors market AI-visibility tools around them. SparkToro's exact 2026 paid-click and Google-properties splits carry the ad-blocker caveat and should be quoted with it, not to a false decimal; the 68.01% zero-click headline is unaffected.

The absence of a confirmable primary source is not evidence that a number is false. It is evidence that the number cannot responsibly be presented as verified. Where a figure resolved, it was reinstated with its source; where it did not, it stays flagged.

Research notes and limitations

This report prioritizes 2024-2026 material and reads first-party Google documentation and Alphabet's own earnings for the claims about ranking factors and market position. Its strongest single source, Pew, is a neutral research organization, but it measured 900 US adults over one month, so it is a strong sample rather than the whole population. SparkToro's multi-year zero-click trend mixes clickstream panels (Jumpshot, Datos, Similarweb), which makes the level for any single year solid and the decade-long comparison directional. Ahrefs and Semrush are search vendors with a commercial interest in the "search is changing" story, so their studies are used for direction with method attached, never as neutral fact. StatCounter measures referral traffic in its network, which sees where clicks land, not where questions are asked, so it undercounts chatbot usage as a research behavior even as it correctly shows chatbots are a tiny referral source. Every click and CTR figure in the category is method-dependent, and this report's central instruction to a founder is the same one it applies to itself: read the method and the date before the number.