Loading page
Target outcome
A category shortlist audit and evidence-backed acquisition plan
Audit the shortlist
Decode the evidence
Locate the gap
Earn the position
Run in Claude and Run in ChatGPT open a new browser tab with the prompt already typed. Nothing is sent until you submit it.
A buyer asks an assistant for the best tool in your category and gets four names back. Either you are one of them or you are not. This workflow spends its calls on a single question: when a direct competitor is named and you are absent, what does that answer cite?
Three plays compete for the same budget every quarter. Buy the review-platform slot. Ship the comparison page. Pitch the trade press. Each is a bet on which kind of source carries your category inside AI answers, and the pages that rank for a query are not necessarily the pages an assistant reaches for.
Your own monitored answers settle that bet, and four calls settle it. So this playbook does not open the overview, the matrix, the competitor tables or the keyword triggers: none of them can change which of the three plays you run.
1. Build the loss list. get_content_opportunities returns two sets: questions where a direct competitor was named and you were not, and questions answered at least twice where you were mentioned under 30% of the time. Both cap at 10 rows whatever limit you pass, and both give question text with no prompt ID.
2. Resolve the text to prompt IDs. get_prompts pages your monitored prompts, up to 100 per page, with their topic and tags. Every row from step 1 is a monitored prompt, so a text that fails to match means you have not paged far enough.
3. Read the losing answers. get_prompt_answers lists recent answers per prompt with provider and whether you were mentioned, truncating each to 500 characters. Use it to choose which answers matter, then get_answer_details for the full text, the competitors named with their position, and every cited source with its domain and rank.
4. Type the cited domains. get_source_domains filters on a domain string and returns the type held for it: institutional, commercial, media, blog, forum, social, educational or other. Use the returned type rather than your own read of the URL.
Tally every citation across the sampled losing answers by the type step 4 returned, then apply this in order.
If one type holds half or more of the citations, run that play this quarter and only that one: commercial for review platforms and marketplaces, media or blog for editorial outreach, forum or social for community presence. If no type reaches half, the sample has not decided, and widening it costs less than splitting a budget three ways.
The comparison-content decision is separate and simpler. If your own domain is cited in none of the sampled losing answers, you have no asset in the room when the question is asked, whatever the type tally said.
Save, per prompt you intend to influence: the period, how many answers you read, whether you were named, and the domains cited. A rerun without that baseline produces a number with nothing behind it.
A competitor's most cited domain is a place to look, not the reason it wins. Nothing here reaches your pipeline, so connect funnel data before claiming revenue.
I want to know which of three plays my category actually supports: review platforms, editorial outreach, or comparison content. Start with list_brands, and if I monitor more than one, ask which before pulling any data. Use the last 30 days unless I say otherwise.
Run exactly these calls and nothing else:
1. get_content_opportunities. Report the missed opportunities, where a direct competitor was named and my brand was absent, and the low visibility questions with their rate. Both lists cap at 10 rows. If both come back empty, tell me plainly that no answer in the window matched the loss test and stop there. Do not substitute brand-wide metrics for the missing evidence.
2. get_prompts, paging until every question text from step 1 is matched to a prompt ID. Report anything still unmatched instead of guessing.
3. get_prompt_answers on the matched prompts. The text comes back truncated, so use it only to choose which answers to open. Then get_answer_details on those, recording the provider, the competitors named with their position, and every cited source domain with its rank. If a matched prompt has no answers in the window, say so and drop it from the tally.
4. get_source_domains once per distinct cited domain, using the domain filter, and report the type it returns. If a domain is not found, count it as untyped and tell me how many untyped citations there are.
Then tally the citations by type and tell me two things: whether any single type holds half or more of them, and whether my own domain was cited in any of them. Give me the counts and the sample size before the conclusion. If no type reaches half, say the evidence does not choose between the plays and tell me roughly how many more answers would need reading. Do not estimate revenue, and do not promise that a published page earns a citation.
Connect Qwairy to Claude, pull the exact signals in the workflow, and leave with an execution-ready output.
Map monitored AI visibility across TOFU, MOFU, and BOFU. Isolate stage and provider gaps, inspect selected answers, and build an evidence-based content backlog.
Use Google Search Console and Bing Webmaster data through Qwairy MCP to discover, review, launch, and measure an evidence-backed AI prompt portfolio.
Audit whether your monitored prompt set gives a balanced view of AI visibility. Find leading language, duplicate intent, coverage gaps, and weak measurement design before updating the portfolio.