You cover what no single rival pass sees: how buyers search and choose in this category, and where our brand stands. Do every section; never stop to ask whether to go deeper. Same rules as the research-competitors-rival-pass skill (../research-competitors-rival-pass/SKILL.md): don't load any skill or start subagents; about 12 searches and 10 fetches in a standard run; public pages only; every fact with its link and date inline; how sure in words; "Needs the browser" and "Couldn't read" lists. The main run has already done the field searches and chosen the rivals — you go deeper on the category: sections 3–7 (and 1–2 only to add what it missed).
Inputs: our brand, what it sells, its category in buyers' words, the rival list, company type, today's date.
1. Money searches
Search, in a clean browser or web search where possible: best <category> · <category> software / tool / app · <rival> alternative (each main rival) · <rival> vs <rival> · <our brand> vs <rival> · <our brand> reviews · our brand name alone. Record who ranks in the top results, who runs search ads on each query (including on our brand name — a rival bidding on us is a threat), and which "vs" and "alternatives" pages exist and who wrote them. Results vary by location; say where you searched from.
2. AI-answer presence
Ask the same prompts each time — best <category> tool, alternatives to <main rival>, <our brand> vs <main rival>, what is <our brand> — in every AI answer engine the host can reach (Google's AI Overview in search results counts; others as available). Record who is named, in what order, and how our brand is described, with the date. Where none can be reached, say so, and read improve-search-visibility's latest search/visibility/<date>.md if it exists.
3. Community — how buyers frame the choice
Last 90 days on Reddit (the category's subreddits), Hacker News, LinkedIn comment threads, Quora: best <category>, <category> recommendation, <category> tool for <job>, switching from <rival>. Themes with links: what buyers compare on, the words they use, which names come up together, and any rival drawing anger.
4. Creator demos and reviews
YouTube, TikTok, LinkedIn and X: <category> tool demo, <category> review, best <category> <this year>, and <rival> demo for any rival launch the rival passes flag. For the 3–5 best (by views or engagement relative to the creator's usual): link, date, creator, length, and the beat breakdown — hook · setup · reveal · proof · CTA · style. These are swipe-file candidates: the format is already proven with this audience.
5. Category-wide ad winners
Ads in the category from anyone, not just named rivals: TikTok Creative Center top ads for the category, Meta and LinkedIn ad-library keyword searches for the category's main terms, Google Transparency for category advertisers. Note long-running ads and repeated formats, and any new entrant advertising heavily.
6. Our own brand in the field
Where our brand shows up in sections 1–5, how it is described, and any recent coverage of our own news (from the main run's news scan) that buyers are reacting to.
7. Adjacent options
For each adjacent option the main run named (a spreadsheet, an agency, an in-house team, a DIY tool): how it is pitched (headline or the way buyers describe it), what it costs, and the top 5 complaints buyers have about it, with links. These are often the real competitor for small buyers.
Return this, in this shape
## Category — as of <date>
Money searches: <query → who ranks · who advertises · comparison pages> (searched from <location>)
AI answers: <prompt → engine → who is named, in order · how we're described> — or "not reachable"
Community themes: <theme — linked examples — names mentioned together>
Creator demos: <link · date · creator · length · beat breakdown> (top 3–5)
Category ad winners: <link · advertiser · start date · format · hook>
New entrants: <name — evidence>
Our brand in the field: <where we appear, how we're described>
Adjacent options: <option — pitch · cost · top 5 complaints (linked)>
Needs the browser: <page — the exact URL or search to open>
Couldn't read: <source — why — what it might have changed>