Primed Prompts

Overview#

Everything else in Seerly runs on a schedule. Primed Prompts is the one place where you can ask a question right now and watch the answer come back while you wait.

It exists for a specific reason. Scheduled monitoring deliberately asks a fixed set of questions so that results stay comparable from one week to the next, which is exactly what you want for trends and exactly what gets in the way the moment you have a hunch about something. Primed Prompts is where the hunch goes.

Direct versus primed#

The idea the screen is built around is the difference between two ways of asking the same thing.

A direct question is a single question asked cold, with no preceding context. It is the closest approximation to somebody opening an assistant and typing one line into it.

A primed or Multi-step prompt is a short conversation that establishes context before arriving at the real question, and Add another primed step is how you build that sequence up.

The comparison between the two is the whole point. A brand can be entirely absent from a cold question and present once the conversation has established what category is being discussed, or occasionally the reverse. Real buyers very rarely ask one question in isolation, so the primed version often reflects a live conversation more faithfully, while the direct version remains a harsher and cleaner test of whether you are known at all. Run Comparison puts the two side by side.

Run Comparison

Asked	Question	Present
Direct	What's the best project management tool for a small agency?	No
Primed	We're a five-person design agency juggling client work.	-
Primed	Which project management tool would suit us?	Yes, second

Running a test#

New Test starts one. Choose an engine, or leave it on All AI engines to run across everything your plan covers. Run Prompts executes the test, and results come back marked Completed or Failed.

Interpreting results#

Present tells you whether your brand appeared at all.

Competitor detection separates direct competitors from indirect ones and from other mentions, and that last bucket rewards reading closely. It catches brands taking the answer slot without being on your tracked list, which is one of the more reliable ways to discover a competitor nobody had been watching.

Results shows the answer text itself, and it is worth reading rather than skimming the flags above it. How you are described matters at least as much as whether you were named, and a mention inside an unflattering sentence counts the same as any other in a summary.

Recent Tests#

Recent Tests keeps your history, so you can return to a run later or compare two phrasings against each other. Back to Recent Tests returns to the list, and Open test to view direct question or Open test to view question flow opens a specific run.

Run Another Test repeats a test, and it is worth doing more often than people do. AI answers vary between runs on their own, so a single result is one data point rather than a finding. If something surprises you, run it twice before you believe it and three times before you tell anybody.

Add to Monitoring#

Add to Monitoring promotes a prompt into your tracked set, and Added to prompt console confirms that it worked. From that point it runs on every scheduled analysis and appears in Visibility like any other prompt.

This is the natural end of a test that turned out to matter. If a question is important enough that you tested it, it is usually important enough to measure properly, and leaving it in the playground means it is only ever answered when somebody remembers to ask.

Best practices#

• A question phrased the way a buyer would ask it, rather than the way you would describe your own product.
• Direct and primed run as a pair, so you learn whether context changes the outcome.
• Repeated at least once before you act on a surprising result.
• Anything that proves important promoted into monitoring, so it starts building a trend instead of sitting in your history.

Next steps#

1. Test the question you are actually worried about. This is the fastest route to answering "what does ChatGPT say about us for this?" without waiting for a scheduled run.
2. Run the comparison rather than only the direct question, so you find out whether the problem is presence or context.
3. Read the indirect mentions for competitors you have not been tracking.
4. Try a rephrasing before concluding you are absent. A question phrased in a way no real buyer would use produces an answer that tells you very little about anything.
5. Add the ones that matter to monitoring and let the scheduled runs build the trend.

Limitations#

A test is a snapshot of one moment. It is not a trend and should not be presented as one, since answers vary between runs even when nothing about your brand has changed.

Tests run against the engines your plan covers. See seerly.app/pricing.

Promoting a prompt to monitoring affects future runs only. It does not backfill history, so a newly added prompt starts its trend from the next analysis rather than filling in the past.