Managing tracked prompts
Every number in OptimizeCamp traces back to a question someone might actually ask an AI. The Prompts screen is where you see that question set, manage it, and watch how each one is doing over time.
What a tracked prompt is
A tracked prompt is a realistic question OptimizeCamp asks each AI platform on your behalf — the kind of thing a real buyer might type into ChatGPT or search on Google. Every metric on the Visibility dashboard is built by running your tracked prompts against every platform, storing the answers, and measuring whether and how your brand shows up in them. For the full mental model of prompts, runs and answers, see Prompts, runs & answers.
The prompts table
The Prompts screen (/[org]/projects/[project]/visibility/prompts) lists every prompt tracked for your project. A subtitle under the heading reads like “24 active · 3 paused — pooled cap 27 / 100 across your projects” (the used/cap turns amber when you’re at the limit). Three status tabs — Active, Paused, Archived, each with a live count — filter the list. Each row shows:
| Column | What it shows |
|---|---|
| Prompt | The exact question text (it’s literally what gets sent to each platform), with a small v2-style badge under it if the prompt has been edited. Click it to open the prompt’s detail page. |
| Category | One of four fixed categories — see below. |
| ChatGPT / Google AI Overview | One column per tracked platform, each holding a status badge for that prompt’s latest run on that platform. |
| Trend | A small sparkline of the prompt’s daily brand-presence score. |
| (actions) | Edit, and Pause / Resume / Archive depending on status (hidden on the Archived tab). |
The per-platform status badge is one of four states:
- No answer — no run has completed for this prompt on this platform yet.
- Absent — the platform answered, but your brand wasn’t mentioned.
- Mentioned — your brand was named in the answer text.
- Mentioned + cited — your brand was named and one of your pages was linked as a source. The strongest outcome.
The four categories
Every tracked prompt is classified into exactly one of four categories. The category shapes how you should read a result — being absent from a branded prompt is a very different problem than being absent from an unbranded one — and you set it when you add or edit a prompt.
Branded
Questions that already name your brand — the buyer knows you exist and is checking you out. Example: “Is Acme good for small teams?” or “Acme pricing.” Losing a branded prompt is the most alarming signal: someone asked about you by name and the AI still didn’t back you up well.
Unbranded
Category-level questions with no brand name at all — the buyer is still shopping. Example: “best project management tools for small teams.” This is the highest-value, hardest-won category: it’s where new customers discover you instead of you having to already be top-of-mind. New prompts default to this category.
Comparison
Head-to-head or alternative-seeking questions. Example: “Acme vs Trello” or “alternatives to Trello.” High-intent — the buyer is actively deciding between named options.
Informational
Educational questions in your topic area with no purchase intent yet. Example: “how to improve team workflows.” These build topical trust over time rather than winning a decision on the spot.
Adding & suggesting prompts
Your first prompt set is generated during onboarding, but the Prompts screen is a full management surface — you add, edit, pause and archive prompts here whenever you like. There are two ways to add:
Write one yourself
Let OptimizeCamp suggest a batch
Suggestions fill real portfolio gaps
Candidate quality is checked first: generic candidates below the relevance floor and reasons that do not name a profile fact are discarded. OptimizeCamp then allocates the four prompt categories against the final portfolio and rotates through the least-covered funnel stages, profile topics, and named competitors. A loss only gets weak-spot priority after it repeats in two independent tracking batches; a single latest absence remains unconfirmed. Each suggestion says exactly which portfolio gap led to its selection. This means “suggest more” complements what you already track instead of returning another arbitrary top-ten list.
On each suggestion row, small chips make that reasoning visible: a funnel-stage chip (awareness, consideration, decision), a category chip, an amber Confirmed loss chip when the gap is a prompt you’ve repeatedly lost, and a one-line reason naming the exact portfolio gap that surfaced it.
The prompt cap is pooled across projects
Your plan’s tracked-prompt limit counts active + paused prompts across every project in your workspace, not per project. Archived and superseded prompts don’t count. If a batch you’re adding would exceed the cap, the whole add is refused with a note telling you how many slots are left — deselect a few or upgrade in Settings.
Editing a prompt & how versioning works
The Edit popover lets you change a prompt’s text or category — but the two are handled very differently, because changing the wording changes what you’re measuring:
- Category-only change: updated in place. The category is just how you read the result, so nothing about the measurement changes.
- Text change: creates a new version. A fresh prompt (v2, v3…) is created carrying over the old one’s status, tags and matched pages, and the old wording is retired to superseded status.
Editing text is non-destructive — but it's a break
The old version keeps all of its runs and answers, so its history stays on the trend chart — nothing is thrown away. But because the question itself changed, a break marker is drawn at the edit date so a shift in the numbers isn’t misread as a real change in visibility. The Edit popover warns you before you save; the two segments either side of the marker aren’t directly comparable.
Editing text is cap-neutral — one active prompt is replaced by one active prompt, so it never uses a slot. A couple of guards apply: archived and superseded prompts can’t be edited, and you can’t edit a prompt into a duplicate of one you already track.
Prompt status
Every prompt carries a status, and it controls participation: only active prompts run when you run a tracking cycle, feed the alert detectors, and appear in the Competitors head-to-head.
- Active — running on the cadence, counts toward your prompt cap.
- Paused — kept but not run; still counts toward the cap. Resume to reactivate.
- Archived — retired by you; doesn’t run and doesn’t count toward the cap.
- Superseded — an earlier version replaced by an edit. Hidden from the list, doesn’t count toward the cap, but its runs still show on the trend.
A project with zero active prompts can’t run tracking at all — the dashboard will tell you so if you try.
The Trend sparkline
The Trend column draws a tiny sparkline of each prompt’s daily brand-presence score — a quick read on whether you’re gaining or losing ground on that specific question. The line is green when the latest point is at or above the first (rising) and amber when it’s falling; until there are at least two runs to plot, it shows a dash (“Not enough runs yet”). Click into the prompt for the full-size version.
The prompt detail page
Clicking a prompt opens its detail screen. Alongside the latest per-platform answers — your brand highlighted green, competitors amber, with citations — see Reading an answer, it adds two things for managed prompts:
- A Presence trend chart spanning every version of the prompt, with break markers at each edit — so a prompt’s whole measured history reads as one honest timeline.
- Version history (shown once a prompt has been edited): every version with its text and date, older ones linkable, so you can open exactly what was being asked at any point. Landing on a superseded version shows a banner linking you to the current one.
Trend breaks, in one place
A trend break is a dashed vertical marker on a trend chart flagging that measurement conditions changed on that date, so a methodology change is never misread as a real visibility change. On a prompt’s chart, a break appears at each text edit (“Prompt edited (v2)”). One other thing draws a break on every chart in the project: changing your brand aliases in your profile, since that changes what counts as a mention everywhere. Category-only edits and other profile edits (description, topics, ICP) deliberately don’t create a break — they don’t change the measurement.
Where your prompts come from
During onboarding OptimizeCamp reads your site and generates a batch of realistic candidates — roughly branded, unbranded, comparison and informational in a sensible mix — grounded in your brand description, audience and topics. You pick which to track. From then on, the Prompts screen (add, Suggest N more, edit) is how the set grows and changes.
Key takeaways
- The Prompts screen is a full manager: add your own, Suggest N more within remaining plan capacity, edit, pause, resume and archive, across Active / Paused / Archived tabs.
- Every prompt falls into one of four fixed categories — branded, unbranded, comparison, informational — each implying a different kind of problem when you're absent.
- Suggested batches balance category allocation and least-covered funnel, topic and competitor exposure across the portfolio; relevance breaks ties.
- Editing a prompt's text is non-destructive but versioned: it creates a new version, retires the old to 'superseded', keeps its history on the chart, and draws a break marker at the edit.
- The prompt cap counts active + paused prompts pooled across all your projects; archived and superseded prompts don't count.
- Each row shows a trend sparkline; the detail page shows a full presence-trend chart across every version, plus version history.
Next step
Click into a single prompt to see what the AI actually said — Reading an answer walks through the brand-highlighted answer text, citations, and which of your pages should be winning that answer.