Is AI Search Killing SEO? What the Verified Data Actually Shows (2026)

Marketers, website owners, and agencies are all suddenly talking about Generative Engine Optimization and Answer Engine Optimization. Before adding to the noise, we did the research — pulled the primary sources, checked the earnings filings, read the independent studies — to separate what’s actually happening from what’s just being sold. Here’s what the numbers say, where this is realistically headed, and how to move with it without overreacting.

The Short Version

Website clicks from search are down — that part is real. But the size and cause of that decline is where most of the noise online gets it wrong. Below, every claim is graded: verified against primary sources, genuinely disputed among credible analysts, or a vendor claim that should be treated with skepticism because the source sells the solution it’s promoting.

What’s Actually Verified: AI Is Changing the Click, Not Killing the Market

Two things can both be true, and the data shows they are:

• [VERIFIED]  Organic search traffic, measured at the market level, is down only modestly. Graphite’s analysis of the top 40,000 U.S. websites (cross-validated against Google Search Console/Analytics with a 0.86 correlation) found organic traffic down 2.5% year-over-year as of January 2026. Google itself stated in August 2025 that total organic click volume has been “relatively stable year-over-year.” This directly contradicts claims of a -25% to -60% market-wide collapse.
• [VERIFIED]  Click-through rate on individual queries drops sharply when an AI Overview appears on that specific query. Multiple independently-run studies converge here: Pew Research (68,000 real searches, 900 U.S. adults) found users clicked through only 8% of the time with an AI summary present, versus 15% without. A randomized field experiment (pre-registered, though still a non-peer-reviewed working paper) found removing AI Overviews increased outbound clicks by roughly 60%, indicating a real causal effect, not just correlation.
• [VERIFIED]  The impact is uneven, not uniform. Graphite’s data shows the largest, most authoritative sites actually grew traffic (~1.6%) while mid-tier sites (roughly ranked 100–10,000) absorbed most of the decline. Informational, low-commercial-intent categories (news, health, entertainment) were hit hardest; e-commerce and transactional queries were far less affected — which is consistent with Google having a direct financial incentive to keep AI summaries away from the queries that monetize best via ads.
• [VERIFIED]  Paid search advertising is not shrinking. Alphabet’s Q1 2026 SEC filing shows total Google advertising revenue grew 15.5% year-over-year, with Search ads specifically up 19%. The only ad segment that declined was Google Network (third-party publisher/app ads), down 4% — a narrow slice, not the whole picture. Any claim that companies are pulling paid ad budget in favor of GEO isn’t supported by Google’s own numbers.

Put plainly: the picture is a redistribution of clicks, concentrated in informational content, not a collapse of search or of paid advertising. Anyone telling you “SEO is dead” or citing a blanket 30-70% traffic loss across the board is overstating a real but narrower phenomenon.

What’s Genuinely Disputed

• [DISPUTED]  How fast this trend will continue. Gartner has been cited projecting a 25% drop in organic search traffic by 2026 and 50% by 2028 — but this is a forecast, not an observed result, and it conflicts with Graphite/Google’s own 2.5% figure for the same period. Projections this far apart tell you the field doesn’t agree yet, and forecasts from firms with a stake in the AI-search narrative should be read cautiously.
• [DISPUTED]  Whether AI Overview CTR damage is worsening or stabilizing. Seer Interactive’s tracking of 53 brands and 5.47 million queries shows AI Overview click-through actually recovered 85% between December 2025 and February 2026 after an earlier steep drop — suggesting users are adapting their behavior (scanning past the summary to find a relevant citation) rather than the damage being permanent or one-directional.

What’s a Vendor Claim, Not a Verified Fact

Almost every statistic claiming GEO/AEO delivers specific outsized conversion lifts (“5x,” “9x,” “27x,” “35% more clicks”) traces back to companies that sell GEO/AEO tools or services — the same conflict of interest that should make anyone skeptical of any industry self-reporting its own category’s success. None of these figures come from an independent, peer-reviewed, or third-party-audited study across a representative sample.

• [VENDOR CLAIM — TREAT WITH CAUTION]  “AI-referred traffic converts at 14.2% vs. 2.8% for Google” and similar 4–27x conversion-lift figures are reported by SEO/GEO tool vendors (Semrush, Ahrefs internal case studies, HOTH, xSeek, ALM Corp) marketing their own GEO services. These may reflect a real effect — users who received an AI recommendation before clicking are plausibly higher-intent — but the specific multipliers vary wildly across sources (2x to 27x) and none are independently verified, which is itself a signal these are selectively reported wins, not audited averages.
• [VENDOR CLAIM — TREAT WITH CAUTION]  Total “GEO market size” figures (e.g. an $848 million to $33.7 billion trajectory cited in some agency material) come from market-sizing firms with a commercial interest in the category’s growth story and should not be treated as settled fact.

The One Piece of Real Academic Evidence

There is genuine, citable research behind part of the GEO practice — just narrower in scope than the marketing claims suggest. A study from researchers at Princeton, Georgia Tech, and IIT Delhi (published as an arXiv preprint, not yet peer-reviewed in a journal) tested specific content changes against AI-generated answers and found that adding statistics, citing credible sources, and improving content fluency measurably increased how often content was surfaced in AI-generated responses — gains in the 15–40% range depending on the technique and query domain.

This is real evidence that certain content practices affect AI visibility. It does not prove those techniques increase revenue, leads, or business outcomes — the study measured visibility in generated answers, not downstream conversions. Treat this as solid but limited: a documented input effect, not a proven business ROI.

Honest Trend Read (Not a Prediction Dressed Up as Fact)

Given the disputes above, the responsible framing for the next 12–24 months is a range, not a confident forecast:

• AI-generated answers will likely keep absorbing a growing share of clicks on informational, research-style queries — this is the one point where independent and vendor data agree directionally, even if they disagree on magnitude.
• Transactional/commercial queries — the ones e-commerce brands care most about — appear comparatively protected for now, plausibly because they’re where Google’s ad revenue is concentrated. That could change if Google’s incentives shift, but there’s no current evidence it’s happening.
• Whether this becomes a 2.5%-a-year gradual shift or a Gartner-style 50% collapse by 2028 is genuinely unknown. Anyone stating either extreme with confidence is going beyond what the data supports.

How to Transition — Without Betting on Unproven Numbers

Given the uncertainty above, the sensible moves are the ones that make sense regardless of how the debate resolves — not a wholesale reallocation of budget based on vendor conversion-lift claims that haven’t been independently verified:

• 1. Fix the technical foundation regardless. Crawlability, valid schema/JSON-LD, and a working llms.txt cost little to get right and help both traditional SEO and AI-visibility outcomes — this is a low-regret move either way.
• 2. Track AI citations as a new, separate metric — not a proxy for revenue yet. Monitor whether and how often your brand appears in ChatGPT, Perplexity, and AI Overview answers. Treat it as a leading indicator to watch, not a KPI with a proven dollar value until you have your own first-party conversion data to back it up.
• 3. Don’t cut paid ad budget on the assumption GEO is replacing it. Alphabet’s own numbers show advertiser demand for paid search is growing, not shrinking. If you’re testing GEO spend, fund it from marginal budget, not by cutting a channel with verified, growing performance.
• 4. Weight any GEO/AEO vendor pitch (including from any agency, including ours) against whether it cites independently verifiable data or its own case studies. Ask for the methodology behind any conversion-lift number before accepting it.
• 5. Re-test your own numbers quarterly. The Seer Interactive data above shows AI Overview CTR patterns changed materially within a two-month window — this space is moving fast enough that a claim from six months ago may already be stale.

Sources Cited

Alphabet Inc. Q1 2026 earnings release (SEC filing, April 29, 2026)

Graphite/Similarweb organic traffic analysis, covered by Search Engine Land (January 2026) ·

Google public statement on organic click volume (August 2025)

Pew Research Center search behavior study (published July 2025, 68,000 searches)

· Seer Interactive AI Overview CTR tracking (53 brands, 5.47M queries, Jan 2025–Feb 2026)

· Aggarwal et al., Generative Engine Optimization, arXiv preprint (Princeton/Georgia Tech/IIT Delhi)

· Various GEO/AEO vendor sources cited explicitly as such and flagged for conflict of interest.