Act on a visibility recommendation
Recommendations turn the prompts where AI answers name somebody else into content, updates, social posts and outreach you can assign and track. Each carries a modeled lift, the buyer questions it must win, and a brief a writer can start from. If you have no finished run yet, start with See whether AI assistants recommend you.
Before you start
Plan required
Below Enterprise the Recommendations tab shows an upgrade panel with Try it free for 7 days. One click, no card, and one trial per feature per workspace, ever. AEO and AEO Advanced are counted separately, so trialling one leaves the other available.
Run a visibility check first. Recommendations are drafted from one finished run's gaps, so there is nothing to draft from until a run has finished and its answers have been saved.
Generate the list
Click Recommendations in the AEO tab row.
Click Generate from latest run.
MarqIQ sends the missed prompts from the latest finished run, the domains the engines cited instead, and the social profiles found on your site, and asks for up to eight items.
Read the amber notice if one appears.
It says what the draft could not cover: prompts no engine answered, a partial run whose gap list is therefore incomplete, and gaps on prompts you no longer track.
Work down the tabs.
- All — everything still open.
- Owned content — net new content and updates to existing pages.
- Social amplification — posts, where your social presence is thin.
- Outreach — the third-party sites the engines cite instead of you.
- Completed — what the team has finished.
Note
Generating again does not replace the list; it adds to it. Prompts an open recommendation already targets are not credited a second time, so the lift column cannot sum past 100% however often you click.
Read a row
| Column | What it holds |
|---|---|
| Title and summary | What to create and why it wins the answers in question. Click the title to open the drawer. |
| Est. lift | The modeled gain in prompt coverage if every prompt this item targets started getting an answer naming you, with the count beneath it: "2 of 40 prompts". |
| Type | Net new content, Update existing, Social amplification or Outreach. |
| Content type | Guide, Educational listicle, Comparison, Case study, Product page, FAQ, and the like. |
| Channel | Blog, Website, LinkedIn, X, YouTube, or a site name for outreach. |
| Priority | high, medium or low, and the order the list is sorted in. |
| Status | Not started, In progress, Completed or Dismissed. |
| Assignee | A free-text name, up to 80 characters. Saved when you click away from the field. |
Three things can appear in the lift column instead of a figure. counted elsewhere means every prompt this item targets is already credited to a higher-priority item or to one already open. not comparable means the row was estimated before lift was measured against your tracked prompts, so its old figure overstated and is not shown. A dash means no prompt was targeted at all — a profile-level item such as social presence.
Take care
The lift is a model, not a promise. It is the share of your active tracked prompts the item addresses, in points of prompt coverage, on the assumption that every one of them starts naming you. Nothing measures whether it did except the next run.
Open the drawer and get the brief
Click the recommendation's title.
Read Suggested content.
Title, summary, and — once a brief is drafted — target audience, primary keyword, three to six secondary keywords, and a four-to-six heading outline.
Click Draft the full brief if the brief is missing.
Rows created before briefs existed have none. One click drafts it on your AI account.
Read What's driving this recommendation.
The buyer questions it targets, each with every engine's verdict in the run it was drafted from: a green ✓ mentions you, a grey — no mention for an answer that did not name you, an amber failed or not asked (quota) where the call produced nothing, and a grey not measured where the fault was MarqIQ's side.
Open the Most-cited URLs in those answers list.
These are the pages the engines leaned on when they answered without you. For an outreach item, they are the pitch list.
Note
The verdicts come from the run the recommendation was drafted from, not from the newest one. Where a newer run exists, the drawer says so and points at the Prompts tab for each question's latest answers.
Work the list
Set Status to In progress when somebody picks the item up.
Type the owner's name into Assignee.
It saves when the field loses focus. A refused change says why rather than reverting in silence.
Set Status to Completed once the work is published.
Completed and dismissed items leave the working tabs; completed ones are on the Completed tab.
Run a new visibility check after the content has been live long enough to be indexed and read.
Keep the prompt set the same across that pair of runs, or the change will not be comparable. Compare runs over time covers the four conditions.
Tip
Do the outreach items alongside the content ones. A page nobody cites is a page the engines have no reason to read, and the cited-URL list is the shortest route to the sites they already trust.
If it did not work
"Run a visibility check first — recommendations come from its gaps."
Why: No finished run exists. A run still being asked, or one discarded unfinished, is not one MarqIQ drafts from.
Fix: Open Overview and click Run visibility check. Resume or discard an unfinished run first if the page offers to.
"The last run's answers could not be saved when it finished."
Why: The answers came back and the write failed after its retries, so there are no gaps left to read.
Fix: Run the visibility check again. The score from that run is still real — it was taken before the save failed — but nothing can be drafted from it.
"The last run got no answers for any of its prompts."
Why: Every call failed, so there is nothing to call a gap and nothing to call a win.
Fix: Fix the engines named on the amber banner on the Overview tab, then run the check again.
"Every answer in the last run is to a prompt that names your brand."
Why: All the active prompts are aided, and an aided prompt is neither a gap nor a win — an answer to a question naming you says nothing about whether AI brings you up.
Fix: Add prompts that do not name you on the Prompts tab, run the check again, then generate.
"Nothing could be drafted."
Why: The drafting model returned nothing usable, or no AI engine could be reached. Anything with a title under 8 characters is dropped as well.
Fix: Check that an engine is connected on the Overview tab, then try again. The message carries the provider's own reason where there was one.
Generating is refused with a wait
Why: Recommendation drafts are limited to 10 an hour and 20 a day per workspace on MarqIQ's included AI engines. Brief drafts have their own budget of 30 an hour and 40 a day.
Fix: The message says when to try again. Connecting your own AI account removes the limit — work on your own account is not counted against it at all.
A row shows no lift although it clearly targets a prompt
Why: That prompt is credited to another open recommendation — a higher-priority one drafted alongside it, or one already in the list.
Fix: Hover the counted elsewhere text. Complete or dismiss the item holding the credit, then generate again if you want the prompt counted here.
A prompt in the drawer has no verdict chips at all
Why: The run this item was drafted from never asked that prompt, or that run is no longer stored.
Fix: Read the line beneath — it says which run the verdicts come from. Open the Prompts tab for each question's latest answers.
Next steps
Related articles
Was this article helpful?