Target outcome
A transcript-optimized video plan grounded in AI citations
Measure citations
Study winners
Map topic gaps
Run in Claude and Run in ChatGPT open a new browser tab with the prompt already typed. Nothing is sent until you submit it.
Connect Qwairy to Claude, pull the exact signals in the workflow, and leave with an execution-ready output.
Plan the videos
Video is the one source type where the domain tells you nothing. An article's domain names its publisher. A watch page's domain is identical whether the video is yours, a competitor's or a reviewer's, so everything decision-grade about video citations lives at URL level. This workflow stays there.
Views and watch time describe an audience on the platform, not whether a watch page was pulled into an answer about your category. Ranking cited domains also breaks here: every channel in your market shares one domain, so its mention count, average position and ownership flag aggregate publishers you do not control.
What the data supports is narrower: a watch URL was cited in a given answer, at a known position, next to known brands. Channel ownership and which passage mattered are not in it, so transcripts stay hypotheses you check by opening the video.
1. Set the baseline. get_social_insights over 30 days returns social citations against total citations, a platform split and the top communities. The split is built from those top community rows, so it can cover only part of the total. Zero social citations ends the workflow here.
2. Find every video host. get_source_domains matches its domain filter as a substring, so a partial search returns desktop, mobile and short-link hosts as separate rows that never merge. Its counts carry no date filter, so read them as all-time. No video host, same stop.
3. Pull the URL inventory. get_source_urls matches a domain exactly, so run it once per host, and it is not period-filtered. get_source_profile is period-scoped and adds unique questions, providers, top URLs and up to ten co-cited competitors. Both group by exact URL, so one video cited with a timestamp arrives as several rows: merge by video identifier first.
4. Tie URLs back to prompts. No source tool returns a prompt per URL. Build the sample with get_prompts and get_prompt_answers, then get_answer_details, the only response carrying the full answer text with source positions. Keep answers whose sources include a video host; if none do, the backlog has no anchor.
5. Check it is not already covered. get_content_opportunities returns up to ten missed questions where direct competitors are named without you, and get_actions returns the live queue, showing whether a topic already has work booked.
A real citation graph put youtube.com second at 3298 citations with average position 8.4, behind a social forum at 4894 citations but at position 9.8.
Fewer citations, better placement. That inversion is the argument for treating video separately instead of folding it into a social bucket. It also sets the ceiling on this workflow: all 3298 rows are the same domain, so nothing in the citation graph separates your channel from anyone else's. The watch-page URLs recovered in step 2 are the only thing that does, which is why the plan is built from them.
Order the shortlist in three tiers. First, topics where a video already appears in an answer to a monitored prompt and that same question also shows up as a missed opportunity. Second, topics carrying a video citation on two or more distinct prompts. Third, everything else, exploratory. Break ties on average source position, then on a co-cited direct competitor, and park any topic backed by a single answer.
Commission one video for the top topic, then stop until the next measurement run: with one you can attribute the change, with five you cannot.
Save the cited URLs with their positions and the answers where a video host and your brand appear together, then re-read on the same prompts and period. Co-citation is an association, and the ownership flag reflects the domain, never the channel.
Start by listing the brands I monitor. If there is more than one, ask which to use before pulling data.
1. Run get_social_insights for the last 30 days. Report social citations, total citations and the platform split, noting the split may not cover the whole social total. If social citations are zero, stop and tell me to re-check after the next monitoring cycle.
2. Run get_source_domains with a partial video domain search so desktop, mobile and short-link hosts all appear. List each separately and present its mention count as all-time. If no video host appears, stop here too.
3. For each host, run get_source_urls on that exact domain, then get_source_profile for the last 30 days. Report URL, title, mentions, average position, unique questions, providers and co-cited competitors, calling co-citation an association rather than an endorsement. Merge URLs that point to the same video but differ by timestamp or playlist parameter. get_source_urls is not period-filtered.
4. Use get_prompts to pick monitored prompts on those topics, then get_prompt_answers and get_answer_details on the latest relevant answers. Keep only answers whose sources include a video host, and show the prompt, provider, full answer, competitors named and the video URL with its position. If none qualifies, stop and say why.
5. Run get_content_opportunities and get_actions. Flag every candidate topic that already appears as a missed question or already has a queued action.
6. Build the backlog in three tiers: a video citation in a monitored answer plus a matching missed question, then a citation across two or more prompts, then exploratory. Break ties on average source position, then on a co-cited direct competitor. Park single-answer topics and recommend one video, for the top tier-one topic only.
Ask me which YouTube channels my brand owns, because nothing in this data identifies channel ownership. Do not invent channel names, view counts, transcript content or a time to first citation.
Build a PR strategy informed by AI citation data. Identify which publications appear most in monitored answers, find competitor coverage gaps, prioritize pitches, and track before-and-after visibility changes following placements.