Most AI visibility tools stop at the diagnosis. HubSpot’s recommendations engine is its attempt to close that gap, turning citation patterns into a prioritized list of things to do, with a projected impact attached to each one.
That is genuinely more useful than a dashboard. It also creates a risk worth naming early: a list with priorities and projections attached feels authoritative, and it is easy to work through it without asking whether these are the right things for your business to do. This article covers how the recommendations are generated, how to read them, how to prioritize properly, and how to tell whether acting on them worked. For the tool overall, see our guide to HubSpot AEO.
Quick Takeaway
- Recommendations are derived from citation patterns, which means they tell you what has earned citations in your category so far. That is a strong signal and a backward-looking one. Use it to close obvious gaps, not as a substitute for deciding what your brand should be known for.
How HubSpot Generates Recommendations
HubSpot AEO recommendations are suggested actions generated from citation patterns across your tracked prompts, each carrying a content type, channel and priority indicator.
HubSpot’s documentation is unusually specific about the method. For each tracked prompt, AEO analyzes answer engine responses and extracts all cited sources, then identifies patterns across those citations to determine the content and structural signals most commonly linked to a citation. The pattern analysis includes frequently cited domains, common content formats, recurring themes and keywords, and competitor presence.
Recommendations are also generated through a site audit, which scans your website directly for content and technical issues.
Knowing the method tells you how to read the output. These are inferences from what has been cited, which makes them reliable for identifying gaps and weak at suggesting anything genuinely new.
The Five Recommendation Views
HubSpot organizes recommendations into tabs:
| Tab | What it covers | Who should own it |
| All recommendations | Everything available | Marketing lead, for triage |
| Site audit | Content and technical issues on your own pages | SEO |
| Owned content | Content to create or update on your site | Content team |
| Social amplification | Engaging with relevant Reddit threads, and creating LinkedIn or YouTube content based on citation trends | Social and executives |
| Outreach | Third-party publishers and sites to approach for inclusion | PR and comms |
Table 1: The recommendation tabs, as described in HubSpot’s documentation.
That table is the most useful thing in this article for planning purposes, because the four action tabs belong to four different teams. A recommendations list handed entirely to the content team will produce movement on one of them.
Reading the Columns: Priority and Citation Lift
Each recommendation carries content type, channel type, recommendation type (for example net new content or update content), priority, status, assignee and citation lift.
HubSpot defines priority as the relative impact of the recommendation based on citation signal strength, and citation lift as the projected increase in your citation rate if you act on it, to be used for comparing recommendations and prioritizing the ones most likely to improve AI visibility.
Two readings worth holding:
- Citation lift is a projection, not a forecast. It estimates effect on citation rate, not on pipeline. A high-lift recommendation on a commercially irrelevant prompt is still commercially irrelevant.
- Priority reflects signal strength, not effort. A high-priority recommendation might be a page update or a six-month PR campaign. The list does not know the difference.
Opening a recommendation gives you the suggested content, including title, summary, target audience and suggested keywords, plus a “what’s driving this recommendation” panel showing associated prompts, content type distribution and top citations identified. Read that panel before acting; it is the evidence, and occasionally it does not support the suggestion as strongly as the priority implies.
The Content and Channel Types You Will See
HubSpot lists eleven content types that may appear: top lists, comparison, product, blog, documentation, wiki, news, PR, research, user reviews and social. And seven channel types: owned, competitor, earned, peer, user-generated content, affiliate and social.
Scanning the distribution of your recommendations by channel is a quick strategic read. A list dominated by earned and user-generated channels is telling you the work sits outside your website, which is the pattern we describe in citation analysis.
The Site Audit: Two Scans Worth Running Early
HubSpot’s site audit organizes recommendations into two types:
- Content: scans your pages against tracked prompts to flag AEO best practice issues, such as missing titles, with results mapped to specific pages.
- Technical: scans for structural and metadata issues, including structured data markup, canonical tags, page speed signals and crawlability, ranked by estimated impact on visibility.
Run both early. These findings are worth fixing regardless of any AI trend, and they are the fastest-moving items on the list. Our technical SEO checklist for AI search covers the same ground independently of the tool.
You select a domain, then click Run content audit or Run technical audit, and can rescan later. Individual recommendations show affected pages, and an Impact tab lets you track a URL, entering the URL and a start date, to see how often the page is referenced by answer engines.
That URL tracking feature is the most underused thing in the product, because it is what lets you attribute a change to a specific page from a specific date.
Generating Blog Posts From Recommendations
For Marketing Hub Professional and Enterprise, HubSpot’s content agent can draft a blog post from an owned-content recommendation. Practical details from the documentation:
- It requires 1,000 HubSpot Credits per blog post, which is a real per-asset cost.
- A Super Admin must opt the account into the relevant beta, and Edit and Publish permissions are needed.
- If brand voice is not set up, you are prompted to add writing samples first.
HubSpot describes the generation process in four phases: research, drawing on brand voice, ICPs, product details, existing blog titles to avoid duplication, recommendation history, keyword data, and web search for pages referencing your brand and competitor positioning; generate, which defines structure and produces title, body and meta description with answer-first sections and FAQ coverage; refine, aligning tone and adding verified internal and external links while excluding competitor domains; and validate, scoring against a quality rubric and checking formatting requirements, with up to two further refinement passes if the draft does not meet the threshold.
That is a more rigorous pipeline than most AI writing tools describe publicly. It is still generating from patterns in what has already been cited, which means it will reliably produce competent category content and will not produce a distinctive point of view. Our piece on Loop Marketing and AI covers where to draw that line.
Prioritizing Properly
HubSpot’s priority and citation lift order the list by likely effect on citations. Add two dimensions the tool cannot see:
| Dimension | Question to ask | Why the tool cannot answer it |
| Commercial relevance | Is the associated prompt one a buyer asks near a decision? | Citation lift measures citations, not pipeline |
| Effort and ownership | Who does this, and how long does it take? | A page update and a PR campaign look identical in the list |
| Durability | Will this still matter in a year? | Pattern analysis is backward-looking |
| Distinctiveness | Does this differentiate us or match the category? | Recommendations describe what already gets cited |
Table 2: What to add to HubSpot’s prioritization.
A workable method: filter to high priority, then keep only recommendations linked to Evaluation and Decision prompts, then sort what remains by effort. Assign each to the right team using the assignee field, and set status as work progresses, since HubSpot notes completed recommendations remain associated with the prompts they affect.
Measuring Results After Optimization
HubSpot is appropriately cautious here: it notes that answer engine responses change over time, so improvements may not appear immediately, and advises reviewing results across multiple analysis cycles.
A workable measurement approach:
- Track the URL you published or updated, with its start date, using the Impact tab.
- Record the change in your own log as well, including outreach placements the tool cannot see.
- Watch the specific prompts the recommendation was associated with, not the overall score.
- Give it multiple cycles before judging, and expect third-party work to take longer than owned content.
- Look at citation rate for the target source type, not just brand visibility, since a citation can be earned before a mention follows.
| Cadence | What you do | Purpose |
| Weekly | Progress statuses and unblock assignees | Keeps the list moving |
| Monthly | Review completed recommendations against their prompts | Early read on impact |
| Quarterly | Re-run the site audit | New technical and content issues |
| Quarterly | Review the channel mix of new recommendations | Whether the work is shifting toward outreach |
Table 3: A continuous optimization rhythm.
THP Studio Perspective
- Our concern with a recommendations engine this well built is that it discourages judgment. The list arrives sorted, prioritized and with a projected lift attached, and working through it feels like strategy. But it derives from what has already earned citations in your category, so following it faithfully produces content that looks like everyone else’s, because everyone else’s is what got cited. Treat it as an excellent gap-closing tool and a poor source of direction, and keep the question of what your brand should be known for well away from it.
Frequently Asked Questions
Key Takeaways
- Read the “what’s driving this recommendation” evidence before acting, since priority reflects citation signal rather than commercial value.
- Split the list across four owners, because owned content, social amplification, outreach and site audit items belong to different teams.
- Run both site audits early; those fixes are worth making regardless of any AI visibility trend.
- Budget for content agent’s 1,000 credits per post if you plan to generate rather than write.
- Treat the engine as a gap-closing tool rather than a source of direction, since it describes what already earns citations.
Work With THP’s Digital & XEO Studio
- THP’s Digital & XEO Studio works across SEO, AEO and AI search visibility for B2B brands. If your recommendations list is growing faster than your team can act on it, prioritizing by commercial value rather than by projected lift is usually the fix.
Author
