SEO Statistics 2026
By Axis Intelligence Research
Co-author: Elena Rodriguez | Last updated: August 15, 2026 | License: CC BY 4.0
When a Google search returns an AI summary, the open web receives 9 clicks per 100 searches instead of 15 — a 40.00% reduction in outbound clicks, measured across 68,879 real searches by Pew Research Center. Over the same period, Google’s Search advertising revenue rose 16.76% year over year to $63.27 billion in a single quarter. Visibility and monetization have decoupled.
Quick Answer
Axis Intelligence Research measures the open web’s retained click value at SVRI™ 72.77 — meaning roughly 73% of the pre-AI organic click economy survives on a search results page carrying an AI summary. The three numbers that define SEO in 2026: traditional-result clicks fall from 15% to 8% of visits when an AI summary appears (Pew Research Center, March 2025 clickstream); AI-summary citations recover just 14.29% of those lost clicks; and Google’s Search advertising revenue still grew 16.76% to $63.27 billion in Q2 2026 (Alphabet, July 22, 2026). Ranking and traffic are no longer the same asset.
Key Findings
- Axis Intelligence Research calculates a SERP Value Retention Index (SVRI™) baseline reading of 72.77 as of the March 2025 Pew measurement window — the share of pre-AI organic click value the open web retains when an AI summary is present.
- Axis Intelligence Research finds that AI-summary citations recover only 14.29% of the clicks displaced from traditional results, derived from Pew Research Center’s measured 15%, 8%, and 1% click rates.
- Axis Intelligence Research finds that Google Search advertising generated 77.51% of all Google advertising revenue and 52.82% of total Alphabet revenue in Q2 2026 — approximately $695.3 million per day — computed from Alphabet’s reported segment figures.
- Axis Intelligence Research finds that AI-summary incidence rises 6.62-fold with query length, from 8% of one- and two-word searches to 53% of searches of ten words or more, per Pew Research Center’s 68,879-query dataset.
- Axis Intelligence Research finds that U.S. workers in the federal occupational category that formally includes search marketing number 899,580 and earn a 28.26% mean-wage premium over all U.S. occupations, per Bureau of Labor Statistics May 2025 data.
How Much Traffic Do Websites Lose to AI Overviews?
The most rigorous measurement of this question is not a vendor study. It is a clickstream panel run by the Pew Research Center, which tracked the actual browsing behaviour of 900 U.S. adults across 68,879 unique Google searches during March 2025. Of those searches, 12,593 returned an AI summary — an incidence of 18.28%.
The click outcomes:
| Condition | Clicked a traditional result | Clicked an AI-summary citation | Total outbound clicks | Source |
|---|---|---|---|---|
| AI summary present | 8% of visits | 1% of visits | 9% of visits | Pew Research Center, Mar 2025 |
| No AI summary | 15% of visits | n/a | 15% of visits | Pew Research Center, Mar 2025 |
| Axis-calculated difference | −46.67% | — | −40.00% | Axis Intelligence Research |
Two derived figures matter more than the headline. First, the citation recovery rate: of the 7 percentage points of traditional-result clicks lost when an AI summary appears, the citations inside that summary return 1 point — a recovery of 14.29%. Being cited in an AI Overview is not compensation for being displaced beneath one. Second, sessions end more often: users abandoned browsing entirely after 26% of pages carrying an AI summary versus 16% of standard results pages.
Elena Rodriguez, SaaS & Business Software Editor: Run this at a realistic scale and the operator maths gets uncomfortable. A site drawing 100,000 monthly search impressions on AI-summary queries books 9,000 sessions where the pre-AI baseline would have delivered 15,000. That is 6,000 sessions to replace, and the replacement channel — being cited in the summary — is worth 1,000 of them. The attribution decks that model “AI visibility” as a substitute channel are quietly assuming a recovery rate around 100%. The measured rate is 14.29%. Every retention curve built on the optimistic assumption breaks in the first quarter it meets real data.
Which query types trigger AI summaries?
Incidence is not uniform, and this is the single most actionable breakdown in the dataset. Pew found AI-summary prevalence scales sharply with query complexity.
| Query characteristic | Share producing an AI summary | Source |
|---|---|---|
| 1–2 word searches | 8% | Pew Research Center, Mar 2025 |
| 10+ word searches | 53% | Pew Research Center, Mar 2025 |
| Begins with a question word (who/what/when/why) | 60% | Pew Research Center, Mar 2025 |
| Contains both a noun and a verb | 36% | Pew Research Center, Mar 2025 |
| All searches in dataset | 18.28% | Pew Research Center, Mar 2025 |
The gradient from 8% to 53% is a 6.62× increase. Long-tail informational content — the exact material most content programmes were built to produce over the past decade — sits in the highest-exposure band. Short navigational and brand queries sit in the lowest.
What do AI summaries actually cite?
| Source type | Share of AI-summary links | Share of standard-result links | Source |
|---|---|---|---|
| Wikipedia, YouTube and Reddit combined | 15% | 17% | Pew Research Center, Mar 2025 |
| .gov websites | 6% | 2% | Pew Research Center, Mar 2025 |
| News websites | 5% | 5% | Pew Research Center, Mar 2025 |
Government sources appear three times more often in AI summaries than in standard results. That is a retrieval-preference signal with direct implications for how content should be sourced: material anchored to primary institutional data is structurally favoured by the systems now mediating discovery.
The typical AI summary ran 67 words (median), with the shortest at 7 words and the longest at 369. Some 88% cited three or more sources; only 1% cited a single source.
The SERP Value Retention Index (SVRI™)
Click rate alone understates the change, because it ignores whether a session continues and how much of the query surface is affected at all. Axis Intelligence Research built the SERP Value Retention Index (SVRI™) — SERP Value Retention Index — to combine those three dimensions into one reproducible number.
What SVRI™ measures: the proportion of pre-AI organic click value the open web retains, expressed on a 0–100 scale where 100 represents full parity with a results page carrying no AI summary.
The formula, fully disclosed:
SVRI = 100 × [ (0.50 × C1) + (0.30 × C2) + (0.20 × C3) ]
C1 Click Retention = (AI-present traditional clicks + AI-citation clicks) ÷ AI-absent clicks
C2 Session Continuation = (1 − AI-present session-end rate) ÷ (1 − AI-absent session-end rate)
C3 Classic Query Exposure = 1 − (searches returning an AI summary ÷ all searches)
Worked calculation, baseline reading:
| Component | Weight | Inputs | Value |
|---|---|---|---|
| C1 Click Retention | 0.50 | (8% + 1%) ÷ 15% | 0.6000 |
| C2 Session Continuation | 0.30 | (1 − 0.26) ÷ (1 − 0.16) | 0.8810 |
| C3 Classic Query Exposure | 0.20 | 1 − (12,593 ÷ 68,879) | 0.8172 |
SVRI = 100 × [(0.50 × 0.6000) + (0.30 × 0.8810) + (0.20 × 0.8172)] = 72.77
All three inputs are drawn from a single primary dataset measured on one panel over one window, which is what makes the components arithmetically compatible. This is the baseline reading, computed against the March 2025 Pew clickstream — the only clickstream-grade dataset that carries all three components from one methodology.
Weight rationale. Click retention carries half the weight because a click is the only component with direct revenue consequence. Session continuation carries 0.30 because an ended session forecloses all downstream discovery, not just the current result. Classic query exposure carries 0.20 because it scales the other two across the query surface without itself measuring behaviour.
SVRI™ sensitivity to AI-summary prevalence
Because C3 is the component most likely to move — Google confirmed at I/O 2026 that AI Overviews passed 2.5 billion monthly active users and AI Mode passed 1 billion — Axis Intelligence Research publishes the index across prevalence scenarios, holding the measured behavioural components constant.
| AI-summary prevalence | SVRI™ reading | Source |
|---|---|---|
| 18.28% (measured baseline) | 72.77 | Axis Intelligence Research |
| 30% | 70.43 | Axis Intelligence Research |
| 50% | 66.43 | Axis Intelligence Research |
| 70% | 62.43 | Axis Intelligence Research |
| 90% | 58.43 | Axis Intelligence Research |
The index floor is high for a reason worth stating plainly: even at 90% AI-summary coverage, SVRI does not collapse toward zero, because clicks and sessions do not go to zero on AI pages — they contract. Coverage published as “AI kills organic traffic” overshoots what any measured dataset supports. What the data shows is compression, not elimination.
Scope of SVRI™. The index reads U.S. desktop-and-mobile browser behaviour on Google only; it does not capture in-app searches, non-Google engines, or query-intent mix by vertical. C1 and C2 are held fixed in the scenario table, so readings above the baseline row model prevalence change alone, not behavioural change. Readings are not comparable across methodology revisions.
Is Google Search Still Growing? The Revenue-Traffic Divergence
Here the data cuts against the prevailing narrative. Every click metric above points down. Google’s search business points sharply up.
| Metric | Q2 2025 | Q2 2026 | Change | Source |
|---|---|---|---|---|
| Google Search & other advertising | $54,190M | $63,271M | +16.76% | Alphabet Q2 2026 results |
| YouTube advertising | $9,796M | $11,055M | +12.85% | Alphabet Q2 2026 results |
| Google Network advertising | $7,354M | $7,303M | −0.69% | Alphabet Q2 2026 results |
| Total Google advertising | $71,340M | $81,629M | +14.42% | Alphabet Q2 2026 results |
| Total Alphabet revenue | $96,428M | $119,796M | +24.23% | Alphabet Q2 2026 results |
Axis Intelligence Research derives three ratios from those reported figures. Search advertising is 77.51% of all Google advertising revenue, 52.82% of total Alphabet revenue, and runs at approximately $695.3 million per day across the 91-day quarter.
Elena Rodriguez: The divergence is the whole story, and it is not a paradox. Google monetises the query, not the click. When an AI summary satisfies the user in place, Google keeps the ad impression and loses nothing; the publisher keeps the crawl and loses the session. Note the one line in the table that fell — Google Network, down 0.69%, the segment that pays other people’s websites. Search up 16.76%, the network that funds third-party inventory down. Those two numbers sit in the same filing and describe the same reallocation. Anyone modelling organic traffic off Google’s revenue growth is reading the wrong side of the ledger.
How Much Do AI Crawlers Take Versus What They Return?
Cloudflare publishes the crawl-to-refer ratio on Cloudflare Radar: HTML page requests from a platform’s crawlers divided by HTML requests carrying that platform’s referrer. A ratio of 5:1 means five pages taken per visitor returned.
For the documented week of June 19–26, 2025, Cloudflare reported ratios spanning from Anthropic’s 70,900:1 down to Mistral’s 0.1:1 — the latter sending ten times as many referrals as crawl requests.
A caution that most coverage of this metric omits, and that Cloudflare states directly: referral counts capture only web-based tools, because traffic from native apps does not send a referrer header. The ratios may therefore overstate the imbalance by an unknown margin.
Axis Intelligence Research declines to publish a single current crawl-to-refer figure, and the reason is itself the finding. The metric is defined over a measurement window, and crawling volumes swing sharply inside those windows: Cloudflare documented Googlebot’s own ratio moving 19.4% in a single week-over-week comparison, driven by a crawling drop that began on one identifiable day. A figure quoted without its window is not a statistic about a platform; it is a snapshot of one week’s crawl scheduling. Widely republished 2026 ratios for the same operator differ by more than an order of magnitude for exactly this reason, and averaging them produces a number with no referent. Practitioners should read the live Radar dashboard for a dated window rather than any figure republished without one.
Who Works in SEO, and What Does the Field Pay?
The U.S. government does not publish an “SEO specialist” occupation. It publishes SOC 13-1161, Market Research Analysts and Marketing Specialists, whose official federal definition includes employing search marketing tactics, analysing web metrics, and developing recommendations to increase search engine ranking and visibility. That is the closest thing to an official headcount for the field.
| Measure | SOC 13-1161 | All U.S. occupations | Source |
|---|---|---|---|
| Employment | 899,580 | 155,495,730 | BLS OEWS, May 2025 |
| Mean annual wage | $89,490 | $69,770 | BLS OEWS, May 2025 |
| Mean hourly wage | $43.03 | $33.54 | BLS OEWS, May 2025 |
| Median hourly wage | $37.87 | $24.51 | BLS OEWS, May 2025 |
Axis Intelligence Research calculates the mean annual wage premium at 28.26% over all U.S. occupations, and the group’s share of total U.S. employment at 0.579%. The OEWS survey covers wage-and-salary jobs only and excludes the self-employed, so freelance and consultant search practitioners fall outside this count entirely — the true practitioner population is larger than 899,580, by an amount the federal data does not measure.
What Is Actually at Stake Commercially?
Organic search sits upstream of a measurable amount of commerce. The U.S. Census Bureau estimated U.S. retail e-commerce at $326.7 billion in Q1 2026 against total retail sales of $1,929.0 billion — a 16.9% share, seasonally adjusted.
| Metric | Q1 2026 | YoY change | Source |
|---|---|---|---|
| U.S. retail e-commerce sales | $326.7B | +9.8% | U.S. Census Bureau, CB26-81 |
| Total U.S. retail sales | $1,929.0B | +3.9% | U.S. Census Bureau, CB26-81 |
| E-commerce share of retail | 16.9% | — | U.S. Census Bureau, CB26-81 |
E-commerce grew 5.9 percentage points faster than total retail. The channel that search feeds is expanding materially faster than the retail economy around it, which is precisely why a 40.00% compression in outbound clicks is a commercial event rather than a technical one.
Elena Rodriguez: Put the two datasets side by side and the strategic conclusion writes itself. The addressable commerce is growing at 9.8%. The click yield per unit of visibility is down 40.00%. A programme that held its rankings flat through 2025 and 2026 is running to stand still, and one that treated ranking position as a proxy for sessions has been measuring an input it no longer controls. The correct denominator changed. Most reporting has not.
Methodology
Collection. Every figure in this dataset was retrieved from a primary source during the production session on August 15, 2026, with source URL and retrieval date recorded in the accompanying CSV. Click-behaviour data comes from Pew Research Center’s metered-browsing analysis of 900 KnowledgePanel Digital participants, covering 68,879 Google searches during March 1–31, 2025, with results pages independently re-collected April 7–17, 2025. Financial figures are taken from Alphabet’s Q2 2026 earnings presentation, published July 22, 2026. Labour figures come from the Bureau of Labor Statistics Occupational Employment and Wage Statistics national release for May 2025. Commerce figures come from the U.S. Census Bureau Quarterly Retail E-Commerce Sales release CB26-81, May 18, 2026. Crawl-to-refer methodology and reference ratios come from Cloudflare’s Radar documentation. Platform user counts come from Google’s published I/O 2026 keynote transcript.
Formulas. SVRI™ is specified in full in the section above, with each component, weight, and input value disclosed. Derived ratios are computed as follows: AI Click Displacement = (AI-absent click rate − AI-present click rate) ÷ AI-absent click rate; Total Outbound Displacement = (15 − 9) ÷ 15; Citation Recovery Rate = AI-citation click rate ÷ (AI-absent click rate − AI-present click rate); Search share of Google advertising = Search & other ÷ (Search & other + YouTube + Network); wage premium = (SOC mean annual ÷ all-occupations mean annual) − 1. All arithmetic was independently recomputed twice, the second pass using exact rational arithmetic rather than floating point.
Scope. The Pew clickstream is U.S.-only, browser-based, Google-only, and dated to March 2025 — flagged as older data because no clickstream study of comparable rigour has been published since, and vendor estimates of AI-summary prevalence vary widely by keyword set. SVRI™ scenario rows above the baseline model prevalence change only. BLS employment excludes self-employed practitioners. Census e-commerce figures use the restated nonemployer-excluded basis and are preliminary for the most recent quarter. Crawl-to-refer ratios are window-specific and are deliberately not aggregated.
Editorial independence. No figure in this dataset was supplied, reviewed, or sponsored by any company named in it.
About This Dataset
Contents: 97 verified observations covering AI-summary click behaviour, query-type incidence, AI-summary citation composition, Google search advertising segment revenue, SEO labour-market employment and wages, U.S. e-commerce commerce volume, and the full component set and scenario table for the SERP Value Retention Index (SVRI™).
Coverage: United States, with global platform user counts where noted. Temporal coverage March 2025 – Q2 2026.
License: CC BY 4.0. You may use, adapt, and republish these figures commercially with attribution to Axis Intelligence Research and a link to this page.
Citation formats:
- APA: Axis Intelligence Research. (2026, August 15). SEO statistics 2026: Click loss, AI search, and what organic visibility is actually worth. Axis Intelligence. https://axis-intelligence.com/seo-statistics/
- MLA: Axis Intelligence Research. “SEO Statistics 2026: Click Loss, AI Search, and What Organic Visibility Is Actually Worth.” Axis Intelligence, 15 Aug. 2026, axis-intelligence.com/seo-statistics/.
- Chicago: Axis Intelligence Research. “SEO Statistics 2026: Click Loss, AI Search, and What Organic Visibility Is Actually Worth.” Axis Intelligence. August 15, 2026. https://axis-intelligence.com/seo-statistics/.
SVRI™ attribution: SERP Value Retention Index (SVRI™) is a proprietary metric of Axis Intelligence Research, first published August 2026. Formula and components released under CC BY 4.0.
Frequently Asked Questions
If AI Overviews cut clicks by 40%, why did Google’s search revenue grow 16.76%?
Because Google monetises the query and the ad impression, not the outbound click to a publisher. An AI summary that resolves the user’s question in place still carries advertising and still counts as search usage. The revenue line and the referral line measure different sides of the same transaction. The corroborating detail sits in Alphabet’s own segment table: Google Network revenue — the segment that pays third-party sites — declined 0.69% in the same quarter Search grew 16.76%.
Should I stop optimising for long-tail informational queries?
The Pew data says those queries carry the highest AI-summary exposure — 53% for searches of ten words or more, against 8% for one- and two-word searches. That is an argument for repricing the expected click yield of that content, not for abandoning it. Long-tail informational pages remain the material AI systems retrieve and cite, and citation carries brand exposure even at a 14.29% click recovery rate. The error is booking long-tail traffic forecasts at pre-2024 click rates.
Does being cited in an AI Overview make up for losing the blue-link click?
Not at the measured rate. Pew found citation clicks occur in 1% of visits to pages with AI summaries, against 7 percentage points of traditional-result clicks lost — a recovery of 14.29%. Citation is worth pursuing for the exposure and for the 88% of summaries that cite three or more sources, but modelling it as a like-for-like replacement channel overstates its traffic value by roughly seven times.
Why does Axis Intelligence Research refuse to publish a current crawl-to-refer ratio?
Because the metric is window-specific and republished values for a single operator differ by more than an order of magnitude depending on the period measured. Cloudflare itself documented Googlebot’s ratio moving 19.4% week over week. Cloudflare also notes that native-app referrals carry no referrer header, so the ratios may overstate the imbalance by an unquantified margin. Publishing one window’s figure as a standing statistic would be misleading, so we publish the methodology and the caveat instead.
How many people work in SEO in the United States?
There is no official SEO occupation code. The closest federal measure is SOC 13-1161, Market Research Analysts and Marketing Specialists, whose definition explicitly covers search marketing tactics and search ranking work: 899,580 wage-and-salary jobs as of May 2025, at a mean annual wage of $89,490. That figure excludes self-employed consultants and freelancers, so it is a floor rather than a full count.
Are .gov and .edu sources genuinely favoured by AI Overviews?
The Pew data supports this for .gov specifically: government sites were 6% of AI-summary links against 2% of standard-result links, a three-fold over-representation. Wikipedia was also somewhat more common in AI summaries than in standard results, while YouTube was somewhat less common. The practical implication is that content anchored to primary institutional data is structurally advantaged in retrieval.
What does an SVRI™ reading of 72.77 mean for my forecasting?
Treat it as a haircut factor on the organic click value of AI-exposed queries relative to a pre-AI baseline, not as a site-level traffic multiplier. It combines click retention (weighted 0.50), session continuation (0.30), and classic query exposure (0.20). If your query mix skews toward long, question-shaped informational searches, your effective exposure is above the 18.28% dataset average, and the 50% or 70% prevalence rows in the scenario table are the more realistic planning anchors.
Is organic search still worth investing in at all?
The commerce it feeds is growing faster than retail overall — U.S. e-commerce rose 9.8% year over year in Q1 2026 against 3.9% for total retail, reaching 16.9% of all retail sales. Even at 90% AI-summary prevalence, SVRI models retained click value at 58.43 rather than near zero. The measured change is compression of yield per unit of visibility, which changes the return calculation and the required volume — it does not remove the channel.
How often will these figures be updated?
Updates are triggered by events, not by a calendar. The specific triggers are a new clickstream study of comparable rigour to Pew’s, Alphabet’s Q3 2026 results, the BLS May 2026 OEWS release, and any material change in Google’s AI Overviews or AI Mode coverage. We do not refresh dates without new data.
Related Axis Intelligence Research datasets: Google searches per day statistics · Google AI Overviews statistics · Digital advertising statistics · Google Chrome statistics