Analyze a Competitor's GEO Strategy in One Conversation

A competitor's mentions climb for weeks and the dashboard tells you it happened, not what to do. This workflow traces that movement to the domains and URLs behind it, ending with a source list ranked by evidence.

The GEO problem it solves

Auditing a competitor's site tells you what they publish, not which pages an engine reaches for when someone asks for the best option in your category. Those pages are frequently not theirs.

Answer evidence moves the analysis onto the domains cited alongside their mentions and the URLs inside them. None of it proves causation: a source profile aggregates a window, narrowing where to look rather than closing the question.

How the workflow breaks down

1. Choose the target. get_competitor_comparison returns your market rank, your mention count and the DIRECT competitors ranked by mentions for the window you request. It carries no history: it picks who to inspect, nothing else.

2. Confirm the movement. get_competitor_evolution returns a daily series of mentions, average position and sentiment. Days without a mention are absent rather than zero: read a gap as a gap.

3. Map the footprint. get_competitor_position returns coverage, share of voice, average position, sentiment, providers, up to ten citing domains, and shopping or local overlap counts. No per-topic split, no first-seen date.

4. Weigh each domain. get_source_profile takes a domain or an ID and returns, for the window, how many distinct monitored prompts cited it, up to twenty of its most cited URLs with average citation position, and the competitors co-cited in the same answers.

5. Read the answers. get_content_opportunities returns at most ten missed questions naming competitors, but no prompt ID, so get_prompts resolves each text to one. get_prompt_answers truncates the text; get_answer_details returns it whole with every cited URL. get_prompt_signals adds four action counts, any of which can be zero.

What to do with the output

Write the snapshot before you touch anything: evolution series, citing domains, a few excerpts, dated. No tool returns a historical source view, so that file is the only baseline a later run has, and whether your response moved anything is next cycle's question.

Then rank the citing domains by how many distinct monitored prompts cited them and work top down. Act on any domain cited in three or more distinct prompts where your brand is absent from its co-cited list: an engine reaches for it repeatedly on questions you care about, and you are not in it. One or two prompts is a watch-list entry, not a project. Ties break on citation count, then on the lower average citation position of the top URL, skipping empty position values.

The Complete Prompt

I want a competitive GEO analysis from my Qwairy data. List the brands I monitor and, if there is more than one, ask which to use. Explain each output before moving on.

1. Run a competitor comparison over the last 30 days: my market rank and mention count, then the DIRECT competitors ranked by mentions. If my own stats come back null, say my brand has no mention in this window and continue on the competitor side.
2. Pull the 30-day evolution of the leading competitor. Days with no mention are missing rather than zero, so describe gaps as gaps and name the first date any sustained movement starts. If the series is empty, say so and continue.
3. Run a position analysis on that competitor: coverage, share of voice, average position, sentiment, providers, citing domains, shopping and local overlap. If no citing domain comes back, say their visibility is not carried by any source we capture and jump to step 5.
4. Run a source profile on its top three citing domains. For each, report how many distinct monitored prompts cited it, its most cited URLs with average citation position, and the competitors co-cited in the same answers. Remind me to check publication dates on the pages, which are not returned.
5. Run content opportunities over 30 days and keep the missed questions naming this competitor. No prompt ID comes back, so resolve those texts with get_prompts, then pull prompt answers and answer details for up to five. Quote how the competitor is framed and what is cited with it. If no missed question names it, say so and move on. Then run prompt signals with limit 10 and report the four counts; attack and defend can both be empty, in which case say the prompt set offers no offensive or defensive target and build from step 4 alone.
6. Close with a ranked source list, not a calendar. Order the citing domains by distinct monitored prompts. Mark act on any cited in three or more prompts where my brand is absent from its co-cited list, watch on the rest, and break ties on citation count, then on the lower average citation position of the top URL, ignoring empty position values. Give each act entry a URL, what is missing, and a note that the effect is untested.