Prompt
The input a user gives an AI model: the question, instruction and any context supplied with it.
In depth
What it really means
A prompt is the whole input, not just the visible question. It includes the system instructions the product sets, the conversation so far, and any documents retrieved and pasted in behind the scenes. What the user types is one part of a larger package.
For marketers prompts matter as market research. The prompts your buyers type are the new keywords, except nobody publishes the volume data. See prompt volume.
How it works
- The user’s text is combined with system instructions and conversation history.
- Retrieved sources may be appended as context.
- The whole package is tokenized and loaded into the context window.
- The model generates a response conditioned on all of it.
Pros & cons
Pros
- Prompts reveal real buyer language in a way keyword tools never did.
- Long, specific prompts carry far clearer intent than two-word queries.
- Collecting them from customers is free and takes one question on a call.
Cons
- No public volume data exists, so you are prioritizing on judgment.
- Prompts vary hugely between people asking the same underlying question.
- You cannot see which prompts your competitors are winning.
Common mistakes
- Assuming prompts look like keywords. They are longer, fuller sentences with constraints attached.
- Building a prompt list internally instead of collecting real ones from buyers.
- Ignoring follow-up prompts, which is where purchase decisions get made.
- Optimizing for prompts about your brand rather than about the problem.
Best practices
- Ask new customers which prompts they used while evaluating, and write the answers down verbatim.
- Mine support tickets and sales call notes for real phrasing.
- Group prompts into problem, comparison and validation stages.
- Use the exact phrasing as headings where it reads naturally.
- Track which prompts return you and which return competitors.
FAQs
What is a prompt?
The input given to an AI model, including the visible question plus system instructions, conversation history and any retrieved context supplied behind the scenes.
Are prompts the new keywords?
Functionally yes, structurally no. Prompts are longer, fuller sentences carrying constraints and context. There is also no volume data, so prioritization relies on customer research instead of tools.
How do I find out what prompts my buyers use?
Ask them. New customers remember what they typed while evaluating. Support tickets and sales call recordings are the next best source.
Should I put prompts in my headings?
Where the phrasing is natural, yes. A heading that matches the question someone actually asks is more likely to be retrieved for it.
Keep reading
Related on LymLyt
Beyond LymLyt
Further reading
Want this working on your site?
We build the content behind the term, ranked in search and cited by AI.
Book a 30-min call →