Skip to main content
Not all prompts are the same. A customer describing a problem — “my skin gets oily by afternoon” — is at a different point from one asking “boat vs noise earbuds.” Cited classifies every tracked prompt by intent, because the same brand can be strong on one kind of question and absent from another, and the work to fix each is different.

​
The six intent archetypes

Cited’s taxonomy describes how a customer frames the question, not which stage of a marketing funnel they are notionally in. That choice is deliberate: the framing is observable in the prompt itself, whereas funnel stage is an inference about someone you cannot see. These are the production tags applied during prompt generation. Every metric in your dashboard can be filtered and broken down by them.
Older brands may still show legacy tags in their data — Category Discovery, Price & Value, Use Case, Gifting, Attribute-Led, Problem-Solving. These predate the six-archetype set and are kept so historical prompts remain readable. New prompts use the six above.

​
Why these, and not the classic search taxonomy

Traditional SEO sorts queries into informational, navigational, commercial and transactional. That taxonomy was built for 2–4 word keyword searches — “HR software India” — where the words themselves barely signal anything and the category has to be inferred. AI prompts are full sentences, and often multi-turn: “What HR software do Indian startups use? Which is best for a 50-person company? How much does it cost?” When someone writes a whole sentence, the framing is right there in the text. A classification that reads the framing tells you more than one that guesses at a funnel stage. It also maps onto work you can actually do. “Commercial intent” does not tell you what to publish. “Budget-anchored” does.
  • More exploratory questions. People use AI to think, not only to look something up. The conversational format invites problem-first framing that a search box discourages.
  • The funnel compresses. A single conversation can move from a problem, to a shortlist, to a head-to-head comparison, to a price question — in minutes rather than weeks.
  • Almost no navigational prompts. Customers rarely ask an AI for a login page. That traffic stays with search engines and bookmarks, which is one reason the classic four-way split fits AI poorly.
  • Research and purchase separate. AI surfaces the options; the purchase happens on the destination site.
Cited’s prompt library reflects this by using conversational, full-sentence prompts rather than keyword-style terms.

​
Reading your own intent split

The useful question is not “which intent should everyone invest in” — it is where does your brand diverge from itself. Measured across tracked brands over 30 days, mention rate on exploratory intents (problem-first, context-specific, feature-curious) and on decision intents (recommendation-seeking, comparison, budget-anchored) are almost perfectly rank-correlated — a brand strong on one is strong on the other. That correlation is mostly just overall visibility, so it tells you little. What varies, and varies a lot, is the gap between them — and it runs in both directions. Two real examples from tracked brands:
  • One brand sits at 2.1% on exploratory prompts and 10.2% on decision prompts. AI will name it when asked to recommend something, but not when a customer describes the problem it solves. Its category presence is thin; it is only reachable once the customer already knows what they want.
  • Another is the reverse: 62.5% exploratory against 46.3% decision. It is present while people are working out what they need, and loses ground exactly when they start comparing options.
Those are different problems with different fixes, and neither is visible in a blended score. Read your own split before deciding where to invest — the direction is a property of your brand and category, not a general rule.
An earlier version of this page claimed that being cited on informational prompts causally lifts share of voice on later commercial prompts. That claim was not supported when tested: the direction of the gap differs by brand, and a cross-sectional measurement cannot establish a “primes the customer for later” effect in the first place. The honest version is the one above — read your own gap.

​
Intent shifts within a conversation

A single conversation can span several archetypes:
  1. “My skin gets oily by afternoon” — problem-first
  2. “What do people use for that” — recommendation-seeking
  3. “Minimalist vs Deconstruct for oily skin” — comparison
  4. “Is there a cheaper option” — budget-anchored
Each step is measured separately, so you can see which step in that chain you drop out of. Dropping out at step 1 and appearing at step 3 is a very different diagnosis from the reverse.

​
How Cited uses intent

  • Every prompt is classified during prompt generation
  • Mention rate, share of voice and average position are all reported per intent
  • Gap analysis surfaces prompts where competitors lead within a specific intent
  • Tasks are intent-aware, because closing a problem-first gap is different content work from closing a comparison gap

​
Frequently asked questions

Because the six describe something observable in the prompt — how the customer framed the question — while the classic four require inferring a funnel stage. On full-sentence prompts the framing is available directly, and it maps more cleanly to what you would publish in response. “Budget-anchored” implies a page; “commercial” does not.
During prompt generation. Cited drafts prompts against target shares for each archetype so a library is not accidentally weighted toward one kind of question, then tags each one. The mix is visible in Prompts & Responses, so you can see whether your library over-represents any archetype.
Rarely realistic. Pick the archetypes where you diverge most from your own average — those are where something specific is wrong, rather than where you are simply as visible as usual.
The classification is platform-agnostic — the same prompt carries the same intent wherever it runs. What differs is platform behaviour within an intent, which is why intent and platform are separate breakdowns rather than one combined view.