Choose the questions to track
The tracked prompts are the questions every visibility check asks, so they decide what your score is a score of. This article covers writing them the way a buyer would, adding and retiring them, the caps, and what changing the set does to your trend. If you have not run a check yet, start with See whether AI assistants recommend you.
Before you start
Set up the brand profile first. Without one there is no brand to detect, so nothing can be labelled aided and nothing can be drafted.
Tip
Click Prompts in the AEO tab row to open the list. The left-hand menu calls the same page Tracked prompts.
Write the question the way a buyer types it
A tracked prompt is a question, not a keyword. It goes to the engines verbatim, with no system framing and no hint that your brand exists. The wording you type therefore decides what you measure.
- Ask at category level: "best marketing operations agency for B2B SaaS", not "marketing ops".
- Use the phrasing your buyers use in a search, a call or a form: "how do I choose", "top tools for", "who does X for Y".
- Put your sales team's real objection-handling questions in. That is usually where the most valuable gaps turn up.
- Keep one question per prompt. Two questions in one line give you an answer you cannot attribute.
Why a question naming your brand does not count
"Is Acme any good?" will always mention Acme. A prompt that contains your brand name, a known alias or your domain is labelled aided, with a one-line reason on the row saying which name it matched.
- An aided prompt is still asked on every run, and its answers are shown in the Answers from its last run column.
- It is left out of the mention rate, coverage, share of voice, sentiment and every breakdown.
- The label is re-read from your current profile on every response, so renaming the brand relabels existing prompts by itself.
- A line above the list counts them: "3 active prompts name your brand, so they are labelled aided: asked on every run, answers shown, left out of every score."
Take care
If every active prompt names your brand, nothing is scored at all. The overview then says that no prompt is scored and links to the Prompts tab.
Draft a starter set
Click Generate prompts.
MarqIQ sends your offerings summary and your existing prompts to an AI with instructions to write natural buyer questions under 140 characters, each with a two-to-four-word topic and no company or brand name in it.
Read the drafted questions that appear.
The unaided rule is enforced in code, not only asked of the model. Anything matching your name, a known alias or your domain is dropped before it is saved, as are duplicates. Ten is the most a single click adds, and fewer appear when the cap leaves less room.
Turn off anything that does not describe a real buying question.
Use the switch in the Tracked column. Only active prompts are asked, and the active count is what the run panel quotes as the next run's size.
Note
Generating needs an AI engine that the workspace can reach. With none available the page says the prompts could not be drafted and asks you to add your own or refine the offerings summary.
Add your own
Type the question into the field at the top of the page.
The field reads Add your own — e.g. best marketing operations agency for B2B SaaS. Repeated spaces are collapsed first. A question under 8 characters is refused. A question over 300 characters is cut to 300 and saved, so shorten a long one yourself before you paste it.
Click Add.
A question already tracked is refused, ignoring case, so a double click or a second tab cannot create a duplicate.
Read the confirmation if the question names you.
It is kept and asked either way, and the message says immediately that it will not be scored: "Added, and labelled aided: …".
Let the gaps suggest the next ones
Suggested prompts come from where your rivals are being named in AI answers and you are not. The evidence is counted from the latest run's stored answers, so each suggestion names the answers it was counted from.
Click Suggest prompts.
MarqIQ finds answers that name or cite a listed competitor and neither name nor cite you, groups them by topic and competitor, and records the domains those answers cited.
Read the Why line under each suggestion.
It is always the counted evidence — "RightWave named in 3 of 4 answers to prompts about marketing automation agencies, none of those naming you · cited: g2.com, clutch.co" — never the drafting model's prose.
Tick the ones worth measuring.
Nothing is added until you tick it. A suggestion naming your brand, a competitor or a cited site, or repeating a prompt you already track, is dropped before you see it.
Click Add selected.
The button counts what you ticked, such as Add 3 selected. Each is added in turn and anything refused is named.
Note
Suggestions need listed competitors, one finished run with its answers still stored, and an AI reader available. When any of those is missing the card says which in words. When the drafting fails but the evidence stands, the gaps are listed on their own under "Where competitors are named and you are not".
How many you can track
| Limit | Value | What it applies to |
|---|---|---|
| Prompts held in a workspace | 200 | Active and inactive together. A workspace already above it keeps its prompts and cannot add more. |
| Prompts added per hour | 60 | Per workspace, on prompts you add yourself and on suggestions you pick. Generate prompts is not counted against it. |
| Active prompts in one run | The whole active set | On your own AI keys. Nothing is sampled or rotated — every active prompt is asked. |
| Active prompts in one run on MarqIQ's included engines | 20 | MarqIQ's included engines are off unless your deployment has them switched on, so this refusal is not one you should meet. If you do meet it, it names how many prompts are active and how many to turn off — and tell us, because it means the included engines are on for you. |
A run over a cap is refused and says by how much. Nothing is ever cut down to a sample.
Twenty to forty questions is a workable set for one business. Every one multiplies by the number of engines, so 40 prompts on 4 engines is 160 calls on your AI account for each check.
Retire a question
There are two ways to stop asking one, and they are not the same.
| Action | How | What happens to the history |
|---|---|---|
| Turn it off | The switch in the Tracked column. | Kept. Its row shows the older run it was last measured in, labelled "From an earlier run" with the date. |
| Remove it | The bin icon at the end of the row, then confirm. | The prompt and its stored answers go together. Past runs keep the score they froze when they finished, and the overview says how many of that run's answers are still stored. |
Tip
Turn a question off rather than removing it whenever you might want it back. A removal is not reversible and takes the answers with it.
Take care
A tracked prompt cannot be edited. To change the wording you remove one and add another, which is a new question with no history — so rewrite a prompt only when you accept losing its past answers and breaking the comparison with the last run.
What changing the set does to your trend
A change between two runs is shown only when both asked the same scored prompts on the same engines and both got every answer back. Adding or retiring a question breaks the first of those conditions.
Click Run visibility check on the Overview tab.
Read the comparability line in the panel before you choose.
It says whether this run can be compared with the last one, and why not when it cannot. A first run, a last run that did not record which prompts it asked, and a changed prompt set are the three reasons, and a changed set is given with the counts added and removed.
Choose one of the two buttons.
- Keep the same prompts and run — protects the trend. When the set already differs from the last run's, this button instead reads Run with the current prompts (not compared with the last run).
- Update the prompts first — opens the Prompts tab with a banner, for when fresher questions matter more than the trend.
Click Run the check with these prompts when you updated the set.
The click is what starts the run, and the run records that you updated the set.
Note
Nothing changes your prompts on your behalf before a run. A run started without that choice is refused, and each run stores which choice was made. You can therefore tell afterwards whether the number moved because the questions did.
If it did not work
The prompt you added is marked aided and you did not expect it
Why: It matched your brand name, a known alias of it, or your domain. The match is on names, not on the literal brand string alone, so "In-House Marketers" and "theinhousemarketers.com" both count.
Fix: Read the reason on the row — it names the term it matched. Rewrite the question at category level and add it again, then remove the aided one if you do not want it asked.
Generate prompts returns nothing
Why: Either no AI engine was reachable, or every draft was dropped for naming your brand or repeating a prompt you already track.
Fix: The message says which. Connect an engine on the Overview tab, or write two or three questions by hand and generate again — the model is told what you already track, so a fuller list steers it away from repeats.
Suggest prompts says there is nothing to suggest
Why: One of three things is missing: listed competitors, a finished run whose answers are still stored, or an AI reader.
Fix: The card names the one that is missing. Add competitors on the Brand profile card, or run a check first. Competitors are counted from the next run, never retroactively.
Adding prompts is refused with a wait
Why: A workspace can add 60 prompts an hour, counting the ones you type and the suggestions you pick. Ticking a long list of suggestions in one go can reach it.
Fix: The message says when to try again. Add the rest after the wait — nothing already added is lost.
You cannot add any more prompts at all
Why: The workspace holds 200, which is the cap. Every prompt held is also sent to the drafting model as "do not repeat", so the set's size is the size of that request.
Fix: Remove the questions you no longer need, then add. Turning a prompt off does not free a slot — an inactive prompt still counts against the 200.
A switch flips back after you toggle it
Why: The prompt was removed in another tab or by a second click, so the change had nothing to apply to.
Fix: Reload the page. The message reads "That prompt no longer exists — refresh the list."
Next steps
Related articles
Was this article helpful?