Visible vs Cited vs Recommended: 3 AI Search Metrics

In AI search, visible means an AI answer mentions your brand. Cited means the answer links to or attributes information to your content as a source. Recommended means the answer actively suggests your brand as a good choice for the user's need. They are three separate outcomes, and a brand can have one without the others.

Most AI visibility reports blend these into a single score, which hides the most useful information. A brand mentioned in 60% of answers but recommended in 5% has a positioning problem. A brand cited constantly for definitions but rarely mentioned by name has a branding problem. A brand recommended often but never cited has a traffic and control problem. This article defines each metric, explains how to measure it and shows which levers improve which outcome.

Why three metrics instead of one

Classic SEO has a clear chain: rank, get clicked, convert. AI answers break that chain. The Pew Research Center found that users clicked a link inside Google's AI summaries on only 1% of visits. So a citation does not guarantee traffic, and a mention without a citation can still influence a decision. Each outcome creates different value, and each fails for different reasons.

What does "visible" mean in AI search?

Visibility is the presence of your brand name in an AI-generated answer, in any context.

Visibility rate is the percentage of sampled answers, for a defined set of prompts, that mention your brand. If you run 200 answers across your prompt set and your brand appears in 70 of them, your visibility rate is 35%.

Visibility is the broadest signal and the easiest to measure. It tells you whether AI systems associate your brand with the topic at all. It says nothing about whether the mention is positive, accurate or influential. A brand can be visible as a cautionary example.

What drives it: broad, consistent brand mentions across the web, a clear entity description and relevance to the prompt's topic. Ahrefs' study of 75,000 brands found branded web mentions correlated with AI Overview visibility more strongly than any other factor tested.

What does "cited" mean in AI search?

A citation is an explicit source attribution, usually a link, that points to your content as support for part of an AI answer.

Citation rate is the percentage of sampled answers that cite at least one of your URLs. Some teams also track citation share, which is your share of all citations in the answers.

Citations can happen without brand visibility. If your blog defines a concept well, an AI model may cite it for that definition while recommending competitors' products in the same answer. Citations also happen through third parties: a review site citing your product is a citation of that site, not of you, but it still shapes the answer.

What drives it: crawler access for AI search bots, pages that answer questions directly, specific evidence such as statistics and sources, and topical relevance. The GEO research paper found that adding citations, quotations and statistics produced the largest improvements in source visibility within generated answers in its benchmark.

What does "recommended" mean in AI search?

A recommendation is an answer that suggests your brand as a suitable choice for the user's stated need, rather than just listing it.

Recommendation share is the percentage of relevant answers in which your brand is actively favored. Measuring it requires classification. A practical three-level scheme:

Recommended: the answer suggests your brand for the user's situation, for example "For a small team that needs X, Brand A is a strong fit."
Listed: the brand appears among options without preference.
Qualified or negative: the brand is mentioned with caveats, for example "Brand A is powerful but expensive for small teams."

Recommendations are the outcome closest to revenue, especially in commercial prompts. They are also the hardest to earn, because the model is weighing you against alternatives using the evidence available about each.

What drives it: a clear statement of who you are best for, specific proof supporting that claim, positive and consistent third-party reviews, and comparison content that states trade-offs honestly.

Side-by-side comparison

	Visible	Cited	Recommended
Definition	Brand is mentioned in the answer	Your content is attributed as a source	Brand is suggested as a good choice
Main question it answers	Do AI systems associate us with this topic?	Do AI systems use our content as evidence?	Do AI systems favor us for the buyer's need?
How to measure	Visibility rate across sampled prompts	Citation rate or citation share	Recommendation share, with sentiment classification
Main drivers	Brand mentions, entity clarity	Crawl access, answer-first content, evidence	Positioning clarity, proof, reviews, comparisons
Business value	Awareness, consideration	Referral traffic, authority, control over facts	Shortlisting, pipeline
Common blind spot	Mentions may be negative or inaccurate	Citations may support competitors' recommendations	Hard to attribute to revenue

How the three metrics interact

Looking at combinations reveals the most. The patterns below are diagnostic heuristics rather than research findings.

High visibility, low recommendation. AI systems know you but do not favor you. Look at how answers describe you. Common causes include vague positioning, outdated reviews, a reputation for being expensive or complex, or competitors with clearer proof for specific use cases.

High citation, low visibility. Your content is useful but not connected to your brand. This is common for companies with strong educational blogs. The model uses your explanation but does not treat you as a vendor in the category. Make sure educational content clearly connects the topic to what you offer, and build brand mentions in category contexts.

High recommendation, low citation. Models favor you based on third-party sources or trained knowledge rather than your own pages. That is valuable but fragile, because you do not control those sources and they may drift out of date. Strengthen your own pages so they are cited alongside third parties.

Low across all three. Start with the basics: crawler access, a clear description of what you do and presence in the sources models cite for your category.

High across all three in one model, low in another. Source preferences differ by model. Profound's analysis found Wikipedia led ChatGPT citations while Reddit led for Perplexity and Google AI Overviews. Compare cited sources across models to find the gap.

How to measure all three consistently

Define a stable prompt set. Group prompts by intent: category discovery, comparison, alternatives, use case, pricing and brand questions.
Run each prompt repeatedly across models. AI answers vary between runs. SparkToro and Gumshoe found the same brand list almost never repeated. Percentages across many runs are more reliable than any single answer.
Record mentions, citations and classification per answer. Store every cited URL so you can see which sources drive each outcome.
Report the three metrics separately, by model and by prompt cluster.
Track accuracy alongside all three. A mention with wrong pricing or outdated features needs fixing regardless of the metric.
Connect to outcomes. Track AI referral traffic, including the utm_source=chatgpt.com parameter that OpenAI says ChatGPT adds to referral links, along with branded search trends and self-reported attribution.

Which metric should you prioritize?

It depends on your stage and goals.

New or unknown brands should prioritize visibility first. If AI systems do not associate you with the category, recommendations are unlikely.
Content-led brands with strong blogs often have citations already and should focus on converting them into visibility and recommendations.
Established brands usually have visibility and should focus on recommendation share and accuracy, especially in comparison and pricing prompts.
Brands dependent on AI referral traffic should watch citations closely, since citations are the main path to clicks.

Common mistakes

Reporting mentions as wins. A mention can be negative. Always classify sentiment.

Counting only your own domain's citations. Third-party citations that discuss you shape answers too. Track them separately.

Measuring once. Single answers are anecdotes. Use repeated runs.

Optimizing only for citations. Being the source for a definition while competitors get recommended is a common and expensive outcome.

A hypothetical example

A hypothetical data analytics startup runs 300 sampled answers across four models. Its visibility rate is 42%, which looks healthy. Its citation rate is 18%, mostly for a popular glossary article. But its recommendation share is only 6%, and most mentions describe it as "suitable for data teams with engineering support." The team realizes its positioning has shifted to non-technical users but its reviews, docs and comparison pages still describe the old audience. The fix is a coordinated update across its site, review profiles and partner listings, plus a use-case page for non-technical teams.

How Bob Builds AI measures the three outcomes

Bob Builds AI's Visibility Monitoring reports visibility rate, citation rate, competitor positioning, sentiment and the supporting sources across AI models. Analytics & Attribution connects recommendation share, citations and AI referrals to conversions, so teams can see which outcome is moving and whether it affects revenue.

FAQ

What is the difference between being mentioned and being cited by AI?

Being mentioned means your brand name appears in an AI answer. Being cited means the answer attributes information to your content, usually with a link. A brand can be mentioned without being cited, for example when the model draws on trained knowledge or third-party pages, and a page can be cited without the brand being recommended.

AI recommendation share is the percentage of relevant AI answers in which a brand is actively suggested as a good choice for the user's need, rather than simply listed. It requires classifying each answer, for example as recommended, listed or qualified, across many sampled runs and models.

Do AI citations drive traffic?

Some do, but less than many expect. Pew Research found users clicked a link inside Google's AI summaries on only 1% of visits. Platforms like Perplexity display citations prominently and can send meaningful referrals. Citations also matter beyond traffic, because they give you some control over the facts AI answers use.

Which is more important: visibility, citations or recommendations?

For commercial outcomes, recommendations are usually most valuable because they shape buyer shortlists. But recommendations depend on visibility and are supported by citations. New brands should build visibility first, content-heavy brands should convert citations into recommendations, and established brands should focus on recommendation share and accuracy.

How do I track whether ChatGPT recommends my brand?

Define prompts buyers realistically ask, run each one multiple times in ChatGPT, and record whether your brand appears, whether it is recommended or merely listed, and which sources are cited. Repeat on a regular schedule. Manual tracking works for a small prompt set, while monitoring platforms automate sampling across models.

Why is my blog cited by AI but my product never recommended?

Your educational content is useful to AI systems, but they may not connect your brand to the product category. Make sure educational pages clearly state what you offer and for whom, publish comparison and use-case content, and build brand mentions in category contexts on review sites and in industry coverage.

Can a negative mention count as AI visibility?

Yes, and that is why visibility alone is misleading. An answer that mentions your brand as expensive, outdated or unsuitable still counts toward a raw visibility rate. Classify sentiment and accuracy alongside visibility so negative or incorrect mentions are flagged for action.

Conclusion

Visible, cited and recommended describe three different relationships between your brand and an AI answer. Visibility shows association, citations show that your content is used as evidence, and recommendations show that the model favors you for a buyer's need. Each responds to different levers, so blending them into one score hides where the real problem lies.

The practical next step is to split your current AI visibility reporting into these three metrics, by model and prompt cluster, and look for the mismatches. They will point directly to whether you need broader brand presence, stronger answer-first content or clearer positioning and proof. Bob Builds AI can help you measure all three and connect them to business outcomes.