What Google's two AI surfaces cite in the US and UK, from 100 queries and 377 answers
Opttab Research. Data collected 9 November 2026.
Google's two AI answer surfaces read the same web and cite it in very different ways. On 9 November 2026 we ran the same 100 English-language queries through Google AI Overviews and Google AI Mode. Sessions were logged out and geo-targeted to the United States and the United Kingdom. The queries cover six intents: best-of shopping, shopping comparison and value, B2B software evaluation, B2B software selection, home how-to, and tech and work how-to. We captured Google's organic top 10 in the same session. Then we matched every cited URL against it and read all 377 answers that came back.
This study builds on two earlier Opttab Research studies that included both Google surfaces. In our Black Friday 2026 study (31 October 2026), Google showed an AI Overview on 83.0% of searches. In our Cyber Monday 2026 study (4 November 2026), it showed one on 91.4%. Here it showed one on 177 of 200 searches. Query sets, intents and markets differ, so treat any comparison with those studies as directional rather than a measured change.
The short version: Google's two AI surfaces cite the web in different ways. AI Overviews almost always cite a page from Google's own top 10 (98.3% of answers with sources). They cite long source lists and lean heavily on YouTube and Reddit. AI Mode cites a top-10 page in 58.3% of answers with sources. It picks the No. 1 result only 33.5% of the time and usually cites one to three sources. On shopping queries it turns to retailers and product cards. In the UK, only 29.0% of citations come from UK sites.
Key findings
Top 10 in, top 10 out. 98.3% of AI Overview answers with sources cite at least one URL from Google's organic top 10. AI Mode answers do so 58.3% of the time. 36.5% of AI Mode answers cite nothing from the top 10.
No. 1 is no guarantee. 75.1% of AI Overview answers cite the page ranked first. Only 33.5% of AI Mode answers do. On best-of shopping queries, 30.6% of answers cite it.
Overviews cite more. 51.5% of AI Mode answers cite one to three sources. Only 3.4% of AI Overview answers stop that short. 48.6% of AI Overview answers cite seven to ten.
YouTube and Reddit lead. 52.5% of answers cite or embed YouTube. On AI Overviews that rises to 80.8%, against 27.5% on AI Mode. Reddit appears in 28.9% of answers and in 44.1% on AI Overviews.
Best-of becomes a shelf. 93.9% of best-of shopping answers show product cards. By exact count, AI Mode showed cards in 43 answers, with 4.67 cards each. AI Overviews showed them in 23.
UK answers borrow sources. Only 29.0% of citations in UK answers come from UK sites. In the US, 52.8% of citations are local. 25.4% of UK answers cite nothing from the UK top 10.
What it means: a top-10 ranking in Google gets you most of the way into AI Overviews. AI Mode needs its own citation plan and its own tracking. For shopping queries, it also needs exact product feeds.
The numbers at a glance
Number | What it measures |
|---|---|
98.3% | of Google AI Overview answers with sources cite a URL from the organic top 10 (n=176) |
58.3% | of Google AI Mode answers with sources cite a URL from the organic top 10 (n=192) |
33.5% | of AI Mode answers cite Google's No. 1 organic result (AI Overviews: 75.1%) |
51.5% | of AI Mode answers cite only one to three sources (AI Overviews: 3.4%) |
80.8% | of AI Overview answers cite or embed YouTube (AI Mode: 27.5%) |
28.9% | of all answers cite Reddit |
93.9% | of best-of shopping answers show product cards or merchant listings (n=49) |
29.0% | of citations in UK answers come from UK sites (US: 52.8%) |
377 | answers read, from 100 queries in two markets on two Google AI surfaces |
Finding 1: AI Overviews cite Google's top 10 almost every time, and AI Mode often doesn't
We matched every cited URL against Google's organic top 10 for the same query, market and session. Among AI Overview answers with sources, 98.3% cited at least one top-10 URL (n=176, 95% CI 95.1–99.4%). Google AI Mode did so in 58.3% (n=192, 95% CI 51.3–65.1%).
Matching by domain instead of URL barely narrows the gap. AI Overviews reach 98.9% and AI Mode 63.5%. Across all 377 answers, 75.6% cited a top-10 URL.
The coded overlap shows the shape of the split. 36.5% of AI Mode answers cite nothing from the top 10. Only 1.1% of AI Overview answers do the same. 64.4% of AI Overview answers draw half or more of their sources from it.
Measure | Google AI Overviews | Google AI Mode | All answers |
|---|---|---|---|
Cite a top-10 URL (computed exactly, answers with sources) | 98.3% (n=176) | 58.3% (n=192) | 77.5% (n=368) |
Cite a top-10 domain (computed exactly, answers with sources) | 98.9% (n=176) | 63.5% (n=192) | 80.4% (n=368) |
Cite nothing from the top 10 (coded) | 1.1% | 36.5% | 19.9% |
Half or more, but not all, of sources in the top 10 (coded) | 64.4% | 23.0% | 42.4% |
Every cited URL in the top 10 (coded) | 9.0% | 22.0% | 15.9% |
All or nothing in AI Mode. AI Mode answers cite only top-10 pages more often than AI Overviews do (22.0% against 9.0%). They are also far more likely to cite none (36.5%). That split fits AI Mode's shorter source lists, covered in Finding 3.
Exact first, coded second. The top-10 shares in the first two rows are computed exactly. Coders agreed on the overlap bands 74.2% of the time, so use those bands for shape rather than precision.
Why it matters: Treat a top-10 ranking as close to a ticket into AI Overviews. AI Mode regularly cites pages from beyond Google's first page, so give it a separate plan with its own tracking.
Finding 2: Ranking first matters far more in AI Overviews than in AI Mode
53.0% of all answers cite the page ranked first in Google's organic results for the same query, market and session (n=377, cluster 95% CI 47.6–58.4%). The surfaces split sharply. 75.1% of AI Overview answers cite it (n=177). 33.5% of AI Mode answers do (n=200).
Intent moves the number too. Home how-to answers cite the No. 1 page 67.7% of the time. Best-of shopping answers cite it 30.6% of the time. US answers cite it more often (56.9%) than UK answers (49.2%).
Group | Answers citing the No. 1 result | n |
|---|---|---|
Home how-to | 67.7% | 68 |
Tech and work how-to | 59.7% | 67 |
B2B software selection | 53.7% | 67 |
B2B software evaluation | 53.1% | 64 |
Shopping comparison and value | 46.8% | 62 |
Best-of shopping | 30.6% | 49 |
United States | 56.9% | 188 |
United Kingdom | 49.2% | 189 |
All answers | 53.0% | 377 |
Read the groups as indicative. Intent and country groups combine both surfaces and hold 49 to 189 answers each.
Why it matters: Keep defending position one, because it keeps you in most AI Overviews. In AI Mode and on best-of shopping queries, the top spot guarantees little. Track which of your pages each answer cites alongside your rank, prompt by prompt, in Opttab's Answer Engine Analytics.
Finding 3: AI Overviews cite more sources than AI Mode, not fewer
We expected AI Mode, the longer conversational answer, to cite more sources. It cites fewer. 51.5% of AI Mode answers cite one to three distinct URLs (n=200). Only 3.4% of AI Overview answers do (n=177).
AI Overviews go wide. 48.6% cite seven to ten sources, and 29.9% cite eleven or more. Across all answers, 29.2% cite seven to ten (n=377, cluster 95% CI 24.7–33.7%). Only 1.9% cite none.
Distinct sources cited | Google AI Overviews (n=177) | Google AI Mode (n=200) | All answers (n=377) |
|---|---|---|---|
None | 0.6% | 3.0% | 1.9% |
1 to 3 | 3.4% | 51.5% | 28.9% |
4 to 6 | 17.5% | 15.5% | 16.4% |
7 to 10 | 48.6% | 12.0% | 29.2% |
11 or more | 29.9% | 18.0% | 23.6% |
Shopping lists run longest. 57.1% of best-of shopping answers cite eleven or more sources (n=49, indicative). Only 10.4% of B2B software selection answers do (n=67).
Why it matters: In AI Overviews you share a long source list with many other sites. In AI Mode each cited slot is scarcer. Know which page holds that slot for every priority query. Opttab's GEO / AEO actions (https://opttab.com/geo-aeo/) map each prompt to your pages and compare them with the most-cited pages.
Finding 4: YouTube and Reddit carry AI Overviews, and AI Mode leans on them far less
52.5% of all answers cite or embed YouTube (n=377, cluster 95% CI 47.5–57.7%). youtube.com is the most-cited domain, in 194 answers. By surface, YouTube appears in 80.8% of AI Overview answers and 27.5% of AI Mode answers. Video is the majority source type in 38.4% of AI Overview answers.
Reddit follows the same pattern. It appears in 28.9% of answers overall (cluster 95% CI 23.6–34.5%). The share is 44.1% on AI Overviews and 15.5% on AI Mode.
Intent shapes the mix. YouTube appears in 76.5% of home how-to answers. It appears in 31.2% of B2B software evaluation answers. Reddit peaks on shopping comparison queries (40.3%) and is lowest on best-of shopping (14.3%).
Group | Cite or embed YouTube | Cite Reddit | n |
|---|---|---|---|
Google AI Overviews | 80.8% | 44.1% | 177 |
Google AI Mode | 27.5% | 15.5% | 200 |
Home how-to | 76.5% | 33.8% | 68 |
Tech and work how-to | 64.2% | 32.8% | 67 |
B2B software selection | 58.2% | 31.3% | 67 |
Shopping comparison and value | 43.5% | 40.3% | 62 |
Best-of shopping | 34.7% | 14.3% | 49 |
B2B software evaluation | 31.2% | 17.2% | 64 |
All answers | 52.5% | 28.9% | 377 |
The most-cited domains. youtube.com appears in 194 answers (51.5%). reddit.com appears in 106 (28.1%), and google.com in 36 (9.6%). Intent groups hold 49 to 68 answers each, so read them as indicative.
Why it matters: For AI Overviews, a useful video and a genuine presence in Reddit threads are part of getting cited. In AI Mode they carry much less weight on their own.
Finding 5: Best-of shopping answers turn into a shelf, and AI Mode stocks more of it
We counted product cards exactly from the collected cards. 66 answers showed them. Google AI Mode showed cards in 43 answers, with 4.67 cards per answer. Google AI Overviews showed them in 23, with 2.91 per answer. The coded share is 17.0% of all answers (n=377, cluster 95% CI 10.8–23.8%). On best-of shopping it rises to 93.9%.
AI Mode also turns to stores for its sources. Retailers and marketplaces are the majority source in 15.5% of AI Mode answers. In AI Overview answers the figure is 1.7%. On best-of queries, 46.9% of answers are retailer-led. 42.9% of best-of answers cite a Google-owned property other than YouTube.
Group | Answers with product cards (exact count) | Cards per answer | Most frequent stores in cards |
|---|---|---|---|
Google AI Mode | 43 | 4.67 | Amazon.co.uk (14), Currys (6), John Lewis (4) |
Google AI Overviews | 23 | 2.91 | Amazon.co.uk seller listings (3), Breville (3), Home Depot (2) |
United Kingdom | 34 | 3.79 | Amazon.co.uk (16), Currys (7), John Lewis (5) |
United States | 32 | 4.34 | Home Depot (5), Best Buy (5), Breville (5) |
All answers | 66 | 4.06 | Amazon.co.uk (16), Currys (7), Home Depot (5) |
Exact counts and coded shares differ slightly. The exact card count finds 66 answers with cards. The coded field finds 64 (17.0%). We lead with the exact count. The best-of group holds 49 answers, so read 93.9% as indicative.
Why it matters: On best-of queries the answer is a shelf of merchant listings, so feed accuracy decides whether your product appears. In AI Mode, retailer pages also win more of the citations. Check that your retail partners carry the same facts you do. Opttab's Feed Optimizer, part of AI Commerce (https://opttab.com/ai-commerce/), checks your catalog quality for ACP and UCP.
Finding 6: UK answers take most of their sources from outside the UK
UK answers produced 785 answer-domain citations. 29.0% of them come from UK sites (95% CI 26.0–32.3%). US answers produced 852 citations, and 52.8% come from American sites (95% CI 49.5–56.1%).
UK answers also stray further from the local results page. 25.4% cite nothing from Google's UK top 10 (n=189). In the US the figure is 14.4% (n=188). Publishers and review sites are the majority source in 33.9% of UK answers. In the US, video leads at 30.3%.
Measure | United States | United Kingdom |
|---|---|---|
Citations from sites in the market asked | 52.8% (n=852) | 29.0% (n=785) |
Answers citing nothing from the local top 10 (coded) | 14.4% | 25.4% |
Answers citing the No. 1 result | 56.9% | 49.2% |
Publishers and review sites as majority source | 21.8% | 33.9% |
Video as majority source | 30.3% | 20.6% |
Answers citing Reddit | 33.0% | 24.9% |
Answers citing or embedding YouTube | 56.4% | 48.7% |
The shelf stays local. Citations travel, but UK product cards still feature Amazon.co.uk, Currys and John Lewis. US cards feature Home Depot, Best Buy and Breville.
Why it matters: UK brands compete with US content for citations in UK answers. Get UK publishers and reviewers to cover you, with UK prices and availability in text. That gives AI a local source to pick. Opttab's Prompt Volume (https://opttab.com/prompt-volume/) shows which countries a prompt's volume comes from, so you can see which UK prompts to work on first.
Also worth knowing
Fresh pages win on software. 83.6% of B2B software selection answers cite a page dated 2026. Only 1.5% of home how-to answers do. Overall, 42.7% of answers cite a 2026-dated page, and 52.2% cite no dated page at all.
Google cites itself on tech and shopping. 14.6% of answers cite a Google-owned property other than YouTube. The share reaches 37.3% of tech and work how-to answers and 42.9% of best-of shopping answers. Home how-to answers show 0.0%. AI Mode does it in 19.5% of answers, and AI Overviews in 9.0%.
Answers name names. 87.3% of answers name at least one specific product, brand or vendor. Every best-of shopping and B2B software selection answer does (100.0%).
AI Mode always answers. AI Mode returned an answer on all 200 searches. Google showed no AI Overview on 23 of 200 searches.
Software vendors get cited directly. Software vendor sites make up 19.0% of the 1,637 answer-domain citations. zapier.com appears in 20 answers, and salesforce.com in 13.
The playbook
Each play below comes from a finding above, with the step to take and how to run it in Opttab. To see where your brand stands today, start with the free AI Visibility Report (https://opttab.com/ai-visibility-report/). It needs no credit card.
1. Audit top-10 positions on your priority queries Evidence: 98.3% of AI Overview answers with sources cite a top-10 URL, and 75.1% cite the No. 1 result. Do this: List the queries that matter in each market. Check where you rank. Fix pages sitting just outside the top 10 first. In Opttab: Prompt Generator (https://opttab.com/prompt-generator/) builds the query set from your topics or your Google Search Console keywords. Answer Engine Analytics (https://opttab.com/ai-visibility/) then shows which of your pages AI cites for each one, by country.
2. Run a separate AI Mode citation track Evidence: AI Mode cites a top-10 URL in 58.3% of answers with sources, the No. 1 page in 33.5%, and only one to three sources in 51.5% of answers. Do this: Track AI Mode citations query by query. Study the cited pages that don't rank. Note what they cover that your page doesn't. In Opttab: Answer Engine Analytics (https://opttab.com/ai-visibility/) shows which pages AI cites for each prompt and market. GEO / AEO actions (https://opttab.com/geo-aeo/) find your content gaps against the most-cited pages and list the fixes.
3. Pair key how-to pages with a video Evidence: YouTube appears in 80.8% of AI Overview answers and 76.5% of home how-to answers; video is the majority source in 38.4% of AI Overview answers. Do this: Publish a short video for each priority how-to page. Keep the steps identical on both. In Opttab: Content Studio (https://opttab.com/generate-content/) writes briefs and outlines from your visibility gaps. Start with the tracked how-to prompts where you aren't cited.
4. Be present where Reddit threads get cited Evidence: Reddit appears in 44.1% of AI Overview answers and 40.3% of shopping comparison answers. Do this: Find the threads cited on your priority queries. Make sure accurate product facts appear there. Take part as a real contributor and leave the promotion out. In Opttab: Answer Engine Analytics shows the pages AI cites for each prompt and market, so the Reddit threads behind each answer are listed with the rest.
5. Keep shopping feeds exact Evidence: 93.9% of best-of shopping answers show product cards, and AI Mode showed 4.67 cards per answer; 46.9% of best-of answers are retailer-led. Do this: Keep merchant feeds, prices and stock in sync with your site. Check that retailer listings carry the same facts. In Opttab: In AI Commerce (https://opttab.com/ai-commerce/), Feed Optimizer checks your catalog quality. The Agentic Commerce Readiness Audit gives you a 0–100 score with a roadmap.
6. Date and refresh comparison pages Evidence: 83.6% of B2B software selection answers and 57.1% of best-of shopping answers cite a page dated 2026. Do this: Show a visible updated date on buying guides and comparison pages. Refresh them when facts change. In Opttab: GEO / AEO actions (https://opttab.com/geo-aeo/) find content gaps against the most-cited pages. Content Studio (https://opttab.com/generate-content/) writes the updates. Agent and website analytics (https://opttab.com/agent-analytics/) shows which pages AI bots read.
7. Give AI a UK source to cite Evidence: Only 29.0% of citations in UK answers come from UK sites, and publishers and review sites lead 33.9% of UK answers. Do this: State UK prices, stock and shipping in text. Brief UK publishers and reviewers with accurate product facts. In Opttab: Run the same prompts in the US and UK in Answer Engine Analytics (https://opttab.com/ai-visibility/) and compare the sources side by side. For daily tracking in both markets, start a 7-day free trial (https://opttab.com/pricing/) or book a demo (https://opttab.com/demo-book/).
Methodology
Queries. We wrote 100 queries in English and ran the same set in two markets, the United States and the United Kingdom. They cover six intents: best-of shopping, shopping comparison and value, B2B software evaluation, B2B software selection, home how-to, and tech and work how-to.
Surfaces. Every query went to Google AI Overviews and Google AI Mode. We collected both planned surfaces. The surfaces show what a searcher sees. This is not a ranking of models.
Collection. We collected data on 9 November 2026 in logged-out sessions, geo-targeted to each country, through the cloro collection API. Google's organic top 10 was captured in the same session. Google showed no AI Overview on 23 searches, so 400 answer slots gave 377 answers with text. Per-answer percentages use these 377.
Exact matching. We computed top-10 citation exactly. Each cited URL and domain was matched against the organic top 10 for the same query, market and session, using the 368 answers with sources. We also counted product cards exactly from the collected cards.
Coding. A model coded all other fields against a written ten-field codebook. It covers source counts, top-10 overlap, the No. 1 result, Reddit, YouTube, Google-owned sources, dominant source type, source freshness, named products and product cards. We also classified 1,637 answer-domain citation pairs covering 729 domains by type and market.
Quality checks. An independent second coder re-coded 120 answers (1,200 judgements). Overall agreement was 88.8%. Dominant source type scored lowest at 60.8% (kappa 0.519). Coded top-10 overlap reached 74.2% (kappa 0.653). Reddit (kappa 0.98) and product cards (kappa 0.968) scored highest, at 99.2% each.
Statistics. Headline figures use 95% Wilson intervals and a cluster bootstrap by prompt.
Limitations
Logged-out only. Signed-in or personalised results may cite different sources.
One day, one run. Answers vary from run to run. Read the figures as a snapshot of one day.
Observational. The study describes what answers cite. It does not show how Google chooses sources or why it picked a page.
No surface-by-intent split. We did not break top-10 citation down by surface and intent together, or AI Overview appearance by intent. So we cannot say whether the surface gap holds within each intent. We report no paired significance test.
Small groups. Intent groups hold 49 to 68 answers each. Product cards appear in 66 answers. Read these as indicative.
Weakest field. Dominant source type had the lowest coder agreement (60.8%). Treat source-type majorities as directional.
Two English-language markets. We covered only the US and UK. Query sets and markets differ from our Black Friday 2026 and Cyber Monday 2026 studies, so comparisons with them are directional.
