Winning YouTube AI Citations for Buy-Intent Search

Definition

A tactic that makes an 8–12 minute YouTube comparison/review video for each pre-purchase search term and structures the query and its answer into the title, chapters and description so it becomes a citation candidate in Google AI Mode. Where Finding Content Gaps from Search Impression Data finds article candidates from existing web impressions, this process splits buy-intent queries into individual videos and tracks whether AI answers actually cite them.

Perspectives

Borja (2026-08-17, X)

Borja studied the answers returned for 120 buying searches and 119 informational searches across 12 categories in US Google AI Mode. YouTube citations appeared in 94 of the buying searches (78%) and 54 of the informational ones (45%), and in the six software categories the buying figure was 51/60. This is only a single snapshot of one country and one Google surface, not causal evidence that publishing a video gets it cited.

Of the 202 cited buying-search videos examined, the median length was 10 minutes 34 seconds, 157 ran over 8 minutes, and 7 were Shorts. 140 had a chapter list, but only 55 had even a single question line in the description. Among the 260 titles cited on buying searches, 227 had a number and 202 had 2025 or 2026 in them. This pattern is a correlation among already-cited videos, not an experiment validating the causal effect of changing titles, length, or chapters.

The recommended sequence is as follows.

1. Pick the top buying keywords of the day someone is choosing, like best, top 10, pricing, cheapest, review.
2. Split the long-tail queries by audience, use case, location and price constraint, and make one video per query.
3. Start the title with the long-tail keyword and add a number, the year and the audience.
4. Record, in one take, an 8–12 minute walkthrough that talks through each product's audience, price and even its downside.
5. If appearing on camera is hard, put a real screen recording behind an AI avatar hook, and limit generated video to b-roll.
6. Add chapters from 0:00 that surface a per-product verdict.
7. Pull question-shaped and eight-word-or-more queries from Google Search Console with a regex: (?i)^(who|what|where|when|why|how|which|can|could|do|does|is|are|should|will|would)\b for questions and (?i)^(\S+\s+){7,}\S+$ for eight words or more. Search Console uses RE2, so lookaheads do not work and matching is case sensitive without (?i). On distribb.io that filter returned 36,200 impressions in three months against 8 clicks — questions Google answered without anyone reaching the site.
8. Put the keyword in the first line of the description, and attach a self-contained 1–2 sentence answer to each real question.
9. On other videos' comments, leave a link only when you've fully answered the question itself, and avoid repeated phrasing.

How to apply

• Fits a product whose buyers search comparison and review terms (best, pricing, review) and a founder who can record an 8–12 minute walkthrough per query; the evidence is one snapshot of US Google AI Mode, so other countries and surfaces are untested.
• Needs no audience, but it does need a site with Search Console impressions to pull real questions for the description — the web-side use of the same export is Finding Content Gaps from Search Impression Data.
• Pick queries from the SaaS SEO Page Type Priority candidates so the page and the video target the same buy-intent term; track discovery in YouTube Studio, web impressions in Google Search Console and the actual AI Mode citation separately — only the last one is success.
• Start with 1–3 long-tail queries and change title, length, chapters and description together; isolating the effect of each element takes later one-at-a-time experiments. Not a short-form tactic.
• If the goal is the product being recommended rather than a video being cited, see Building Product Information Surfaces for AI Search. Distribb is the author's own product offered as the automated version — judge it apart from the study.

Limits

• The sample is one point in US Google AI Mode and may not reproduce when the country, language, or search surface changes.
• It is an observational study counting the traits of already-cited videos, so it cannot remove survivorship bias and confounders.
• 94/120, the numbers and years in titles, and the video-length and chapter ratios are the original author's published tallies; the raw data and reproduction procedure were not independently verified in this ingest.
• Comment links must answer the question directly and avoid duplicate or bulk posting. The YouTube policy collected alongside the original explains that it prohibits repetitive, untargeted comment spam and the mass production of synthetic content.

Original link