Long Google Queries Drew 37% Forum and Video Links vs 15% for Head Terms

Ras Al Khaimah, United Arab Emirates, 1 November 2026.

Google says people now type far longer questions into AI Mode, and our own test suggests the length of a query also changes what sits on the regular results page. When we ran 10 topics through Google at five lengths, from one-word head terms to 20-plus-word questions, forum, video and social results made up 37% of the organic links for the longest versions, against 15% for the shortest. AI Overviews appeared on 48 of the 50 searches, whatever the length.

The test follows Google's Ads Decoded episode published on 28 October 2026, in which Ads Product Liaison Ginny Marvin and Shivani Mohan, Search Vice President of Data Science and User Experience Research, discussed how AI is changing search behaviour. The episode description says people are "asking 3x longer questions". That figure comes from Google's US AI Mode insights report, which states that the average AI Mode query is triple the length of a traditional search query and that more than one in six AI Mode searches are multimodal.

Google has not published how query length affects the classic results page. We pulled 50 live US desktop results pages on 1 November 2026 to find out what a searcher sees when they type an AI Mode style question into the normal search box.

Key findings

• AI Overviews were near universal. 48 of 50 searches showed one. The two without were the head term "car battery" and the 12-word "what is the best laptop for 4k video editing under 1500 dollars".
• Longer queries pulled in more community content. The share of organic links pointing to Reddit, YouTube, Quora, Facebook, Instagram or TikTok, or displayed by Google with a comment, answer, view or reaction count in place of a web address, rose with length (chart below).
• The "Discussions and forums" panel was a long-query feature. It appeared on none of the 10 shortest searches and on 5 of 10 in each of the two longest tiers.
• Long AI Overviews arrived differently. All 10 of the longest queries returned an AI Overview that our SERP data feed flagged as loading asynchronously, meaning Google filled it in after the first page response. Only 1 of 9 head-term overviews did.
• More words did not mean more sources. Overviews cited a mean of 8.6 to 10.9 unique pages per tier, with no upward trend.

How the results page changed by length

Each topic was written five ways. "Cholesterol" became "lower ldl cholesterol", then "how to lower ldl cholesterol naturally", then a full question about avoiding statins, and finally a 22-word message that included the searcher's own LDL reading. The tiers averaged 1.7, 3.5, 7.2, 13.8 and 24 words, so the longest tiers run many times the length of the head terms, the kind of gap Google describes between AI Mode and classic searches.

Tier (mean words)	AI Overview	Discussions and forums panel	People also ask	Related searches	Image pack	Things to know	Organic links per page
1 (1.7)	9 of 10	0	7	9	5	5	7.9
2 (3.5)	10 of 10	1	7	10	3	2	8.0
3 (7.2)	10 of 10	1	9	10	0	0	8.0
4 (13.8)	9 of 10	5	9	7	0	0	7.9
5 (24.0)	10 of 10	5	9	5	1	0	7.3

Counts are results pages out of 10 per tier. Source: SaaS Link Building analysis of Google US desktop results, 1 November 2026.

Two patterns stand out. Browsing features built for broad intent, such as image packs and the "Things to know" module, all but disappeared once queries passed about seven words. Related searches also thinned out, from 10 of 10 pages in tier two to 5 of 10 in tier five, which fits a query that already says what the searcher wants.

The organic list shifted toward first-hand experience. For the 23-word Japan rail pass question, four of the eight organic links were Reddit threads or Facebook group posts. For the 27-word laptop question, every one of the nine organic links was a forum, video or social post. Even topics with strong publisher coverage moved: the longest security deposit query still led with legal aid and court self-help pages, but the long car battery query opened with a Reddit thread.

That fits how Google describes AI search behaviour. Its report highlights rising follow-up questions, and the Ads Decoded episode description talks about people adding personal nuance and context. We have seen the same community pull elsewhere: YouTube was cited in every one of 30 AI Overviews in an earlier pull, and Reddit's dependence on search traffic was clear when it announced a logged-in only test after losing logged-out users.

What the asynchronous overviews mean

Our data feed flags an AI Overview as asynchronous when it was not in the first page response and had to be requested separately. Twenty of the 48 overviews worked this way, and they clustered at the long end: 1 of 9 in tier one, 1 of 10 in tier two, 2 of 10 in tier three, 6 of 9 in tier four and 10 of 10 in tier five. All five security deposit queries also loaded this way at every length, the only topic to do so.

The reason is not documented, and Google has not commented. Our reading, which is speculation, is that longer or more personal questions need the overview to be generated at request time rather than served from a cache, which takes longer than the rest of the page. One practical effect is on measurement: in asynchronous overviews the top-level source list held just two or three links, while the overview text itself cited 5 to 14 pages. Tools that count only the headline source list will undercount citations for long queries.

Context

This is the length dimension of a question we tested earlier by intent. In our 150-search study of AI Overviews by query type, overviews showed on 77% of searches but rarely on local ones. This time every query was informational or commercial, which helps explain the 96% rate. Separately, an earlier 155-site Search Console study found AI Mode style queries growing sharply in site owners' own reports, a sign that these longer questions are already reaching the regular search box.

Google's insights report is also careful about scope. Its methodology note says its keyword data comes from Google Trends samples between AI Mode's US launch in May 2025 and April 2026, and the length comparison is an average. It does not say how long queries typed into the regular search box behave.

What this means for SEO teams

• Audit long-tail SERPs by hand. Pull the results for the 15 to 25 word questions your customers ask in sales calls and support tickets, not just your head terms. In our test those pages looked different from the short-query versions of the same topic.
• Expect community competitors. For long, situational questions, the pages ranking beside you may be Reddit threads, YouTube videos and Facebook groups. Content that adds specific, first-hand detail (numbers, setups, failure cases) answers the same need those threads meet.
• Write for the situation, not the keyword. Long queries carried context such as a budget, a reading, a deadline or a failed fix. Pages that address those conditions directly have more to match against.
• Check how your tools count AI Overview citations. If a tracker reads only the top-level source list, asynchronous overviews will look thinner than they are.

Methodology

On 1 November 2026 we wrote 10 topics (health, retirement accounts, travel, laptops, plumbing, CRM software, baking, running, renting and car repair) at five lengths each, from a one to three word head term to a 22 to 27 word question in the first person, for 50 queries. We requested each query once from a SERP data API for Google, United States, English, desktop, first page, with asynchronous AI Overviews loaded. Four requests returned a server error on the first attempt and succeeded on retry. For each page we recorded the result types present, the number of organic links, each organic link's host (taken from the breadcrumb when Google masked the link), whether an AI Overview appeared, whether it was flagged as asynchronous, and the unique URLs it cited. "Forum, video and social" counts organic links on reddit.com, youtube.com, quora.com, facebook.com, instagram.com or tiktok.com, plus links Google displayed with only a comment, answer, view, like or reaction count, which in our data were mostly posts on those platforms. Community subdomains of other sites were not counted. This is a single snapshot from one location and device with a hand-picked query set; results vary by user, time and location, and the sample is too small for statistical significance.

Conclusion

Google says AI Mode questions are three times longer than classic searches. In our test, typing that kind of question into the normal search box still produced an AI Overview almost every time, but the page around it changed: fewer browsing features, a forums panel on half the longest searches, and more than a third of organic links going to community platforms.

For teams planning content around long, conversational queries, the competition on those results pages is often a forum thread rather than a rival brand. The pages that hold up are the ones that answer the specific situation in the query.

Questions and answers

How much longer are AI Mode queries than regular Google searches?

Google's US AI Mode insights report says the average AI Mode query is triple the length of a traditional search query. The report's keyword data comes from Google Trends samples between AI Mode's US launch in May 2025 and April 2026.

Did longer queries trigger more AI Overviews in this test?

No clear difference. AI Overviews appeared on 48 of the 50 searches across all five length tiers. The two searches without one were "car battery" and a 12-word laptop query.

What changed on the results page as queries got longer?

Forum, video and social links rose from 15% of organic results for the shortest tier to 37% for the longest. The Discussions and forums panel appeared on 5 of 10 pages in each of the two longest tiers and none of the shortest, while image packs and the Things to know module all but disappeared.

What is an asynchronous AI Overview?

In our SERP data, an AI Overview is flagged as asynchronous when it was not in the first page response and had to be requested separately. All 10 of the longest queries returned one, against 1 of 9 head-term overviews. Google has not documented why.

###