Prompt set design
Build a mix of branded, unbranded, comparison, recommendation and buyer-intent prompts tied to the actual customer journey.
Grow Marketing Hub benchmarks a brand across defined branded and unbranded prompts, source and citation patterns, competitor mentions, factual accuracy and available first party platform data. The goal is a repeatable baseline, not a collection of random screenshots.
We start with the current search baseline, prioritize the highest-impact gaps and validate what changes after implementation.
AI output can vary by model, prompt, date, location, session and product changes. We record the exact test conditions and separate observed answer-set positions from official metrics such as Bing AI citation data.
The exact scope depends on the website, platform, current performance and business priority.
Build a mix of branded, unbranded, comparison, recommendation and buyer-intent prompts tied to the actual customer journey.
Check whether the system can find the brand by name and separately whether it surfaces the brand when the user asks only about the category or need.
Record whether the brand appears and which pages or sources are referenced when the platform exposes source information.
Compare which competitors appear, how they are described and which sources seem to support their visibility.
Check names, products, services, locations, pricing context, availability and other decision-critical facts for missing or incorrect information.
Evaluate whether a user can confidently visit, compare, book, contact or buy after receiving the answer.
The strongest fit is where this service solves a real search, site or growth problem rather than being added as a generic SEO task.
Create a baseline so later changes can be compared against the same prompt framework.
The business wants to know whether relevant AI tools mention it for non-branded category or recommendation prompts.
Competitors appear frequently in AI answers and the business needs to understand the source and evidence gap.
Customers use AI to compare products, services, locations or providers before taking a commercial action.
We group prompts by intent such as discovery, comparison, recommendation, product or service fit and branded verification. A fixed set makes later checks more meaningful because the conditions are consistent.
We can observe what is surfaced, which sources are visible and where the website is weak. We do not claim certainty about why a model chose one brand over another when the platform does not provide that explanation.
The audit can include selected AI-assisted search and answer platforms that are relevant and accessible at the time of testing. Because product availability and output change, the final report records the specific platforms and dates used.
There is no universal number. The prompt set should be large enough to represent the important customer journeys without creating hundreds of redundant variations. Branded, unbranded, comparison and high-intent prompts are usually separated.
Not necessarily. Composite scores can be useful internal summaries, but they should not be confused with a universal platform ranking. We preserve the underlying prompt, mention, citation and source evidence.
Yes where a platform exposes repeatable citation data or where a defined prompt benchmark can be rerun. Bing Webmaster Tools, for example, now provides AI citation and grounding-query data in public preview.
A system may retrieve a brand correctly when given the exact name while never recommending or mentioning it for broader category questions. Separating those tests gives a more useful picture of real discovery.
Share your website and goal. We will review where visibility, page intent, technical issues or conversion paths are most likely holding growth back.