Core concepts

Prompts, runs & answers

Every chart, score, and alert in OptimizeCamp traces back to three simple objects: a prompt you track, a run that asks it, and an answer that comes back. Here's how they fit together.

Prompt: the question you track

Prompt

A tracked question — the literal text sent to each answer engine, on a repeating basis, so you can watch how the answer changes over time. Every prompt has:

A category — branded, unbranded, comparison, or informational (see the table below).
Free-form tags layered on top of the category, for your own organization.
A status — active, paused, archived, or superseded (an earlier version retired by an edit). Only active prompts run; active and paused both count toward your pooled prompt limit, while archived and superseded don’t.
A version — editing a prompt’s wording keeps the old text’s history and starts a new version, so changing the question never quietly rewrites the past. See editing & versioning.
A source — how it entered your set: generated during onboarding, proposed by “Suggest 10 more,” or written by you.

The four prompt categories

Every prompt falls into one of four shapes, each testing a different kind of question a real buyer might ask:

Category	What it tests	Example
Branded	Do AI answers know and vouch for you specifically?	“Is {brand} good for {ICP}?”
Unbranded	Do you show up when nobody asked for you by name — the hardest, most valuable visibility to earn.	“best {category} tools for {ICP}”
Comparison	How do you fare head-to-head against a named rival?	“{brand} vs {competitor}”
Informational	Do you get cited as a helpful source on topics in your space, even with no brand name in the question at all?	“how to get started with {topic}”

Run: one prompt, one platform, one moment

Run

One prompt executed on one platform at one point in time. A run tracks its own lifecycle — queued, running, succeeded, or failed — and belongs to exactly one batch (below). A run’s only job is to call the platform and hand back whatever it returned; everything else happens right after, on the answer.

Answer: what actually got captured

Answer

The result of a run: whether the platform actually responded, the full answer text, the ordered citation list, and metadata about how it was produced. Every answer is parsed immediately for brand mentions and citation ownership — see Mentions, citations & share of voice for exactly how that works. There’s exactly one answer per run.

How a batch works

Runs don’t fire one at a time in isolation — they’re grouped into a batch: every active prompt, times every tracked platform, executed together as one unit.

Guard and kick off

OptimizeCamp refuses to start a second batch while one is already in flight for your project, and marks any batch that’s been stuck for too long as failed so a new run can start cleanly. A fresh batch begins with a job count equal to your active prompts times your tracked platforms.

Fan out

Every (prompt, platform) pair becomes one job. Jobs run a handful at a time rather than all at once — a real web-search answer can take several seconds to come back, so a controlled amount of concurrency keeps the batch moving without overwhelming anything.

Capture and parse each answer

For every job: call the platform, store the raw answer, run entity matching and citation classification against it, and judge sentiment. Each job succeeds or fails independently, so one bad answer doesn’t take down the rest of the batch.

Close the loop

Once every job finishes, the batch triggers the rest of the product: recomputing which pages should be fixed next, checking for content gaps, verifying any pending Impact Cards, refreshing the day’s metrics, and running alert detection — all from the run that just completed.

Run cadence & credits

How often your prompts run automatically depends on your plan. On top of that automatic cadence, every plan also includes a monthly allowance of on-demand runs for when you don’t want to wait — say, right after you’ve published a fix and want to see it reflected today.

Plan	Automatic cadence	On-demand “Run now”
Starter	Weekly (both platforms)	2/month
Pro	Google AI daily · ChatGPT 3×/week	10/month
Agency	Google AI daily · ChatGPT 3×/week	40/month

Every batch is tagged with what triggered it — a scheduled run, a manual “Run now,” or a one-off like the onboarding baseline run. Scheduled runs happen automatically on your plan’s cadence and are never charged against your on-demand allowance; only a manual “Run now” draws from it. Your very first full sweep — the baseline run that produces your first report at the end of onboarding — doesn’t count against it either.

How runs roll up into daily metrics

Once a batch finishes, OptimizeCamp recalculates that day’s Mention Rate, Citation Rate, Position Score, Sentiment Index, Visibility Score, and Share of Voice — for your brand and for every competitor you track — from every answered run captured that day. That recompute is idempotent: rerunning it for the same day always produces the same result, so re-parsing history after a detection improvement, or backfilling a day that predates a feature, is always safe. See Metrics & the Visibility Score for what each of those metrics means.

Key takeaways

A prompt is a tracked question with a category, tags, and a status; a run is one execution of it on one platform; an answer is what came back.
Runs are grouped into batches — every active prompt times every tracked platform, executed together — which is also what triggers priority, gap, impact, and alert recomputation once it finishes.
Your plan sets an automatic per-platform cadence (weekly on Starter; Google AI daily and ChatGPT 3×/week on Pro and Agency) plus a separate monthly allowance of on-demand 'Run now' runs; scheduled and baseline runs never draw down that allowance.
Every finished batch recomputes that day's metrics for your brand and every competitor, using the same rules every time.

Next step

All of this — mention detection, aliases, prompt suggestions — is shaped by one thing you set up first: Your brand profile.