To measure earned media visibility in AI answers, start with a stable set of buyer-relevant prompts, observe those prompts across selected answer engines, and keep brand mentions separate from source citations. Then compare cited pages with a separate PR placement inventory before deciding whether a placement was actually cited.
This is an additional visibility signal, not a replacement for coverage quality, audience reach, referral traffic, or business outcomes. The measurement becomes defensible when the team preserves four distinct data grains: one prompt-engine observation, one citation, one earned-media placement, and one period-level aggregate.
HubSpot AEO is documented as a beta feature that tracks prompts across ChatGPT, Gemini, and Perplexity. Availability and limits depend on the subscription and permissions, so confirm the current setup and reporting documentation before designing an operating process around it.
A citation proves that a source appeared in a tracked answer. It does not, by itself, identify a PR placement as the cause of brand visibility.
What should a PR team measure?
A practical AEO measurement system answers different questions with different fields. Do not collapse them into one count called “AI coverage.”
- Brand visibility: HubSpot defines this as the percentage of tracked prompts in which the brand appears. The denominator is the configured prompt set, engines, competitors, and reporting period, not every question asked online.
- Brand mention: The answer includes the brand. This can happen without a cited source that names or belongs to the brand.
- Citation: The answer references a source page or domain. A citation can appear without the answer mentioning the brand, and the cited source may be a competitor or a general educational page.
- Share of voice: An aggregate proportion of tracked brand mentions belonging to the configured brand among tracked competitors. It is not a metric for an individual citation or placement.
HubSpot’s citation analysis documentation describes reporting by cited domains and pages, channels, content types, and citation-related measures. Preserve separate fields for brand appearance in the answer, whether a source was cited, and whether the cited page itself mentions the brand, if that last point has been verified.
A cited domain is a discovery signal, not a placement match. Store the cited URL and page evidence first, then test it against the PR inventory. This prevents a publisher-level report from being presented as article-level attribution.
Build a repeatable observation baseline
Before interpreting trends, configure the brand name and spelling variations, owned domain, competitors, and buyer-relevant prompts. HubSpot documents that brand variations are case-sensitive, so check spelling and capitalization carefully. Keep the prompt wording, prompt-set version, engine, competitor configuration, and reporting window with every review.
HubSpot documents daily prompt tracking and advises reviewing results over multiple days or weeks because answers can change. Keep daily observations intact and create weekly or monthly summaries as separate records. Compare like with like: the same prompts, engines, competitor set, and date-window rules. When prompts change, create a new prompt-set version rather than silently changing the denominator. Teams reviewing configuration and operational fit may also consider HubSpot systems consulting.
The following is an illustrative measurement-ledger design, not a guaranteed HubSpot export schema:
{
"record_type": "citation",
"observation_id": "OBS-2026-10-09-GEMINI-P014-090000Z",
"citation_id": "CIT-OBS-2026-10-09-GEMINI-P014-090000Z-01",
"engine": "Gemini",
"prompt_id": "P-014",
"prompt_version": "v3",
"observed_at": "2026-10-09T09:00:00Z",
"cited_url": "https://publisher.example/category-guide",
"brand_mentioned_in_answer": true,
"brand_mentioned_in_cited_source": false,
"match_status": "manual-review"
}One observation record represents one identified prompt-engine run. One citation record represents one cited source within that run, so multiple citations from the same answer receive separate citation identities. A placement record describes a PR item, while an aggregate record describes a period-level metric such as visibility or share of voice. Keep these records separate.
If a separate database stores the ledger, define a uniqueness constraint for the intended observation and citation identity, and use an atomic upsert or transaction when concurrent writers are possible. A read-then-insert check alone is not race-safe. Do not use a lossy key such as date plus domain, because the same domain can supply several pages, engines, or citations in one period.
Match cited sources to actual PR placements
Start with the cited URL, not just the publisher domain. One publisher may host unrelated articles, syndicated copies, press-release mirrors, and multiple versions of a page. A domain match is not enough to label a specific placement as cited. Keep the PR placement inventory as its own system of record.
This is a proposed operating workflow, not a documented HubSpot feature. Use deterministic checks for URL identity, redirects, publication dates, campaign IDs, and duplicate handling. AI may assist with an ambiguous content-type label or summarize why a page is relevant to a prompt, but it should not decide URL identity, deduplicate records, validate permissions, or make an irreversible inventory update.
Use citation patterns to improve PR decisions
Review which domains, channels, and content types appear in tracked responses, then compare those patterns with the media list. HubSpot AEO recommendations are based on observed citation patterns and may identify content or channel opportunities. Treat estimated impact as a recommendation to evaluate, not a promised visibility gain.
Prioritize an outlet only after reviewing the relevant cited pages and prompt context. A frequently cited source may deserve closer review, but its appearance in one category or prompt set does not predict performance in another. Consider audience fit, editorial relevance, relationship status, and campaign goals alongside AEO observations.
| Trigger | AI job | Validation | Action or fallback |
|---|---|---|---|
| Unclear page format | Suggest a content-type label | Reviewer checks page metadata and visible content | Save approved label; route uncertainty to manual review |
| Possible placement match | No URL-identity decision | Compare canonical URL, article identity, and date | Update PR evidence only after exact or reviewed match |
| Period closes | Summarize validated changes | Check scope, denominators, and prompt-set version | Publish an observed association or hold the report |
Use approved domain lists and page metadata before asking a model to interpret an ambiguous source. Validate any suggested label against allowed values such as news article, comparison, review, or documentation. If the page is syndicated or its canonical URL cannot be verified, retain the citation as unresolved rather than forcing it into the placement inventory.
Report trends without claiming causal PR attribution
Compare campaign dates with visibility, citation, and share-of-voice trends while showing the engine, prompt-set version, competitor configuration, and reporting window. The result is observational: a metric change after a campaign does not establish that a particular placement caused it. Prompt changes, answer-engine variation, publication timing, concurrent campaigns, and other content activity may also matter.
A reporting record should identify its scope and evidence. Include the reporting period, prompt-set version, engines, competitor configuration, observation dates, metric definitions, cited URLs where available, source evidence, and reviewer status. Keep the interpretation labeled as an observed association.
HubSpot’s connector documentation describes using supported AI assistants to review AEO performance, manage prompts, and review or act on recommendations, subject to permissions. The reviewed materials do not verify a public AEO API, bulk citation export schema, webhook, or native PR-platform synchronization. Before proposing a CRM sync, confirm the actual access and capture path. Keep the placement inventory as the system of record unless a supported integration is independently confirmed. For teams deciding where validated evidence belongs, CRM systems consulting can help assess system-of-record and workflow design.
- Confirm the prompt-set version, engine set, competitor configuration, and date range.
- Check that the denominator is consistent with the comparison period and disclose any scope change.
- Keep aggregate visibility and share of voice separate from individual citations and placements.
- Retain cited URLs, observation times, source evidence, and match-review status where available.
- Confirm that any CRM or campaign update has a validated source identity and an accountable reviewer.
- Use observational wording unless the analysis establishes a stronger design than timing and correlation.
A defensible AEO report makes its observation scope easy to inspect. It helps PR teams investigate which sources appear in answer-engine responses while keeping citations distinct from placements, visibility distinct from mentions, and measured association distinct from causal business impact.