A sweep is one complete measurement pass: every active prompt in your brand's panel is run against every tracked engine — ChatGPT, Gemini, Perplexity, and Claude — and every answer is judged and stored. Sweeps are the heartbeat of Brandflare; everything on your dashboard is an aggregation of sweep results.
Scheduling and cadence
Sweep frequency is set by plan (see plans and limits):
- Daily on Trial, Pro, and Agency: your brand is swept every day.
- Weekly on Starter: your brand is swept once a week, on a fixed weekday assigned deterministically per brand. Staggering brands across weekdays keeps results flowing steadily rather than piling up on Mondays — but your brand's day never changes, so week-over-week comparisons stay clean.
You can also trigger a sweep on demand from the brand page. Manual sweeps always run in full, regardless of schedule — useful right after onboarding, or after you've shipped something you want to measure the effect of.
What happens in a sweep
For each prompt-engine pair, the pipeline does three things.
1. Ask. The prompt is sent to the engine exactly as a user would ask it, and the full response text is captured verbatim. When an engine grounds its answer in web sources, the cited URLs are captured too — over time this builds a picture of which domains feed the answers in your category.
2. Judge. The raw answer is unstructured prose, so scoring it is itself an AI task. A frontier-model LLM-as-judge evaluates each answer for:
- Mentions — whether your brand is named (including registered aliases and product names), which competitors are named, each mention's rank within the answer, its sentiment, and its strength.
- Accuracy — every claim the engine makes about your brand is checked against your fact sheet; affirmative false claims become severity-graded flags. Omissions are never flagged. Details: accuracy monitoring and flags.
- Narrative — on plans with narrative analysis, a written read of how the answer framed your brand.
3. Store. The response, sources, mentions, and flags are all persisted against the run. Nothing is discarded: any number on your dashboard can be traced back to the exact answers behind it.
Runs execute in bounded parallel batches, so a full sweep — panel size times four engines — completes in minutes. Individual runs can fail (engines have outages and rate limits); failed runs are recorded as errors and excluded from metrics rather than counted as non-mentions.
Aggregation
After the runs finish, the day's results roll up into per-brand, per-engine daily metrics:
- Mention rate — the fraction of the day's answers that named your brand.
- Average rank — where you appeared among the brands named, when you appeared.
- Share of voice — your share of all brand mentions (yours plus competitors') in the day's answers.
- Average sentiment — how favorably you were framed.
- Open flags — the count of unresolved false-claim flags from the day's runs.
These daily metrics are what the dashboard charts, what the AI Visibility Score averages over its trailing window, and what the prompt matrix breaks down cell by cell.
Alerts
Once aggregation completes, alert emails go out to workspaces on plans that include them: if the sweep surfaced new high-severity flags — an engine asserting something false about your brand — you hear about it the same day, not at your next login.
Why scheduled sweeps beat spot checks
Asking ChatGPT about your brand once tells you almost nothing: engines are non-deterministic, and the same prompt yields different answers run to run. Sweeps solve this with volume and repetition — a full panel, four engines, on a fixed cadence. Single-answer noise averages out, and what remains is signal: real movement in how AI presents your brand, attributable to a date, and therefore to whatever you shipped that week. That's the discipline of prompt tracking, automated.
What a sweep costs you
Nothing per-run — sweeps are included in your plan. The plan's prompt limit and sweep frequency are the levers that determine how much measurement you're getting: Starter's 15 prompts weekly is a pulse check; 40 prompts daily on Pro and Agency is a monitoring system. If you're deciding between them, the question is how quickly you need to detect change.