A prompt asking an AI system to explain a concept can often be answered without naming a single company. A prompt asking for the best option, a head-to-head vs. comparison, or a recommendation cannot.
That difference in task design helps explain one of the clearest findings in Semrush and Kevin Indig's 2026 research: comparative prompts were associated with substantially more brand mentions than informational prompts. For marketers, the implication is not to abandon educational content. It is to recognize that brand visibility is highly dependent on the kind of decision the prompt asks AI to make.
“Best,” “Vs,” and “Recommend” prompts create more brand-naming opportunities because the answer must identify, compare, or choose among market participants.
This is an interpretation of observed query behavior, not proof that inserting one of these words into a prompt mechanically causes a brand mention.
The Semrush/Kevin Indig study found a 2.4× mention gap
In a June 2026 study conducted with Kevin Indig and Growth Memo, Semrush analyzed 3,981 domain appearances across 115 prompts, 14 countries, and four AI search experiences: ChatGPT, Google AI Overviews, Gemini, and Google AI Mode.
The researchers tracked whether each domain appearance was cited as a source and whether the associated brand was actually named in the answer. When the results were segmented by intent, comparative prompts stood out.
Brand-mention rate for comparative prompts such as “best,” “vs,” and “recommend.”
Brand-mention rate for informational prompts such as “what is,” “explain,” and “how does.”
Approximate increase in brand mentions for comparative prompts versus informational prompts in the dataset.
Informational prompts also had an 89.3% citation rate. That combination is revealing: AI systems frequently used domains as sources for explanatory questions without naming the associated companies. Comparative prompts changed the job from explain this to evaluate the players.
The 2.4× figure is not evidence that comparative prompts are universally better. It shows that, in this dataset, prompts requiring evaluation of alternatives were much more likely to contain brand names than prompts requiring explanation. That is exactly what we should expect if the answer cannot complete the task without identifying options.
Why “Best” prompts naturally create a brand shortlist
A “best” prompt usually asks AI to reduce a market into a manageable set of candidates. “What is CRM software?” can be answered with a definition. “What are the best CRM platforms for a 20-person sales team?” requires named products.
The system now has to perform at least three additional jobs: determine which brands belong in the category, decide which criteria matter, and explain why certain options fit the user's constraints better than others. Brand names stop being optional examples and become part of the answer's structure.
This is why “Best X for Y” queries can be particularly valuable. The “Y” introduces a qualifying dimension such as industry, company size, budget, geography, use case, or feature requirement. The AI is no longer looking for a generic market leader. It is trying to match brands to a specific buyer context.
Why “Vs” prompts force direct entity comparison
A “vs” prompt starts with the brands already inside the consideration set. The user's question might be “HubSpot vs. Salesforce for a mid-market sales team” or “Agency A vs. Agency B for enterprise SEO.” The answer cannot avoid the named entities because comparing them is the entire assignment.
That makes “vs” visibility strategically different from broad category visibility. The goal is no longer to be discovered from scratch. It is to be described accurately on criteria that matter to a buyer: pricing model, feature depth, specialization, implementation burden, customer fit, service model, integrations, or another decision factor.
For brands, this creates a positioning problem as much as a visibility problem. Appearing in the answer is not enough if the comparison consistently frames a competitor as the better fit for your target segment.
Why “Recommend” prompts turn information into a choice
Recommendation prompts often combine discovery and evaluation in the same request. “Recommend an SEO agency for a multi-location healthcare organization” asks the AI to identify possible providers, interpret the user's constraints, and decide which ones deserve consideration.
The answer may still include caveats and alternatives, but it generally needs named entities to be useful. That is the commercial value of recommendation visibility: the brand is being inserted into the buyer's decision process rather than merely supplying background information.
Semrush's 2026 consumer research adds context for why that matters. In its survey of more than 1,000 U.S. consumers, 43% said they had discovered a new brand through an AI tool, while 47% said they noticed AI-mentioned brands often or very often. The study also found that clear descriptions, value context, fit for specific needs, and direct comparisons were stronger attention drivers than simply appearing earlier in an answer.
The three prompt types create different brand-mention jobs
| Prompt type | What the user is asking AI to do | Why brands get named | Content that can support the answer |
|---|---|---|---|
| Best | Construct a shortlist or rank viable options. | The answer needs category participants and criteria for inclusion. | Best Of lists, expert roundups, buyer guides, use-case recommendations. |
| Vs | Compare named alternatives on relevant decision factors. | The named entities are the subject of the answer itself. | Comparison pages, reviews, decision matrices, competitor analyses. |
| Recommend | Match options to the buyer's stated need. | The answer must identify viable choices and explain fit. | Buyer guides, “Best For” content, reviews, expert recommendations, case-oriented content. |
Comparative intent changes what evidence AI needs
An informational answer can rely heavily on facts. A comparative answer needs relational evidence: how one option differs from another, which audience each serves, what strengths and limitations matter, and under what conditions one choice becomes more appropriate.
That means the visibility strategy must extend beyond publishing more pages about your own product or service. Owned content remains essential for accurate facts, capabilities, positioning, case studies, and conversion. But comparative answers can also draw on reviews, publisher lists, directories, community discussions, analyst coverage, and other third-party sources that place the brand inside a market context.
The strongest recommendation footprint therefore tends to include both first-party truth and third-party comparison context. The first explains what the brand is. The second helps establish where it belongs relative to alternatives.
Prompt wording matters beyond intent labels
The Semrush/Kevin Indig study found that prompt length also changed mention behavior dramatically. Short, conversational prompts in its dataset produced much higher mention rates than long, highly structured prompts, even when they addressed similar subject matter.
That is an important warning for AI visibility programs. Tracking one perfectly engineered prompt for each topic can create a false sense of precision. Real buyers may phrase the same need as “best accounting software for contractors,” “what accounting app should I use as a contractor?” or “QuickBooks vs. Xero for a small construction company?”
Semrush's current prompt-research guidance recommends focusing on the middle- and bottom-of-funnel questions where AI evaluates options and makes recommendations, while also varying phrasing and constraints rather than assuming one prompt represents the entire demand pattern.
Do not optimize for the words “best,” “vs,” or “recommend.” Optimize for the commercial decision behind them.
Prompt wording is a signal of buyer intent. The real opportunity is to understand which choices the buyer is asking AI to make, what criteria the answer needs, and what evidence makes your brand a credible participant in that decision.
How to build a comparative-prompt visibility strategy
Map the commercial decisions that matter
Start with buyer questions rather than a giant prompt list. Identify the category, “Best For,” competitor, alternatives, pricing, use-case, and recommendation decisions most likely to influence a real purchase.
Track prompt families, not isolated prompts
For each decision, test multiple phrasings and constraints across the AI platforms that matter. Measure how often your brand, competitors, and cited sources change across the family.
Separate mention, citation, and recommendation outcomes
A cited page is not the same as a named brand, and a named brand is not automatically a recommendation. Track the three outcomes independently so you know which visibility problem you actually have.
Audit the sources shaping comparative answers
Look for the recurring publishers, reviews, lists, comparison pages, community sources, and directories AI uses when it names competitors. Those sources reveal where the market's recommendation context is being formed.
Build evidence for specific fit
Strengthen the factual basis for why your brand belongs in a shortlist: audience fit, capabilities, differentiation, case evidence, limitations, pricing context, geography, or other decision criteria. Avoid fabricated “best” claims.
Close third-party recommendation gaps
Prioritize credible third-party content types that match the missing decision context—Best Of lists, comparisons, reviews, buyer guides, expert roundups, alternatives pages, use-case recommendations, directories, or decision matrices.
What not to conclude from the 2.4× finding
The statistic is useful precisely because it shows how intent changes answer behavior. It becomes less useful when marketers turn it into a shortcut.
- It does not mean informational content is low value. Informational content can earn citations, support topical authority, answer early-stage questions, rank in traditional search, and establish expertise.
- It does not mean every “best” prompt will mention more brands. Results vary by topic, model, country, phrasing, and the evidence available to the system.
- It does not mean publishing a page titled “Best X” guarantees a mention. AI systems can use many source types, and third-party evidence may be especially important when the task is comparative.
- It does not prove causation. The study observed higher mention rates for comparative prompts. The most reasonable explanation is that comparative tasks require named options, but the research does not isolate a single causal mechanism.
Measure the part of AI visibility that matches the buyer moment
If the business goal is authority on an informational topic, citation visibility may matter significantly. If the goal is to enter a commercial shortlist, brand mentions and recommendation frequency become much more important.
A useful dashboard should therefore segment performance by prompt intent and report at least four separate signals: mention rate, citation rate, recommendation frequency, and competitive share of voice. It should also show which sources are repeatedly shaping the answer.
That is the bridge between the research and execution. The Semrush/Kevin Indig finding is not simply that three prompt words are powerful. It is that AI visibility changes when the user's task changes from learning to choosing.
Build coverage around the prompts where buyers are choosing
Mention Engine helps brands identify and build credible third-party recommendation content around the commercial questions that matter: Best Of lists, comparisons, reviews, buyer guides, expert roundups, use-case recommendations, directories, and other decision-oriented formats.
The goal is not to manufacture superiority. It is to make sure credible evidence for your brand exists in the places and formats AI systems may use when buyers ask which options deserve consideration.
Explore Mention Engine