JournoFinder's AI assistant is only as good as the brief you give it. Whether you're using the in-app search/list builder or driving it through the MCP (via Claude or another AI client), the assistant works from the same underlying engine - so the same principles apply.

This guide covers how to brief it well, with good-vs-bad examples throughout.

The best mental model is a capable colleague who just joined your team. They know how to find journalists and read coverage, but they don't know your campaign, your client, or what "good" looks like to you. Everything obvious in your head needs to be said out loud.

That means four things are worth stating explicitly every time: who you're pitching for, the angle or story, the kind of journalist you want, and any hard constraints (geography, outlet type, beat). Get those four in and the assistant has enough to work with.

A topic is a noun. A story is something a journalist would actually write about. The assistant ranks journalists by how closely their recent coverage matches what you describe, so the more you describe an actual story angle, the better the match.

Weak: "Find me journalists who write about electric vehicles."

Strong: "I'm pitching a study showing EV charging deserts in rural areas are widening the urban-rural divide. I need journalists who cover transport policy, infrastructure, or rural affairs - ideally ones who've written about the practical realities of EV ownership."

The second version tells the assistant the shape of the story (data-led, policy-adjacent, rural framing), which surfaces reporters who actually cover that beat rather than every person who's ever typed the word "EV."

If something is non-negotiable, say so explicitly - don't assume the assistant will infer it. Geography is the most common one people forget. "Manchester housing" and "housing, Manchester-local outlets only" produce very different lists; the first may pull in national reporters who covered a Manchester story once, the second keeps it to journalists at Manchester outlets.

Weak: "Journalists covering housing in Manchester."

Strong: "Journalists covering housing, restricted to Manchester-local and regional outlets only - no national desks. Prioritise property, local government, and community beats."

The same goes for outlet tier (trade vs. consumer vs. national), language, and beat. If you only want broadcast, say broadcast.

When the assistant knows who you're pitching and why, it can judge relevance better and you can reuse that context across follow-ups.

Weak: "Sleep journalists."

Strong: "I'm building a list for a consumer-health brand whose spokesperson is a sleep scientist. We've got a study on how screen time in the hour before bed affects sleep quality, based on data from 10,000 participants. I want reporters who cover health and wellbeing and would quote an expert."

This gives the assistant something to rank against ("would this person quote a sleep scientist?") rather than a bare keyword.

The assistant works best when you aim to build one list per request. If you want to build multiple lists, plan to use several requests and not lump multiple lists into one request.

One list can be built from several search angles.

There are two ways to do this, and both work:

1) Add the angles one at a time (recommended for quality). Start with your strongest angle, review what it returns, then add the next angle during refinement.

This lets you check each angle as it lands and adjust before adding more.

Example: "Build a list of wellbeing reporters for a sleep study, focused on the headline finding that screen time before bed lowers sleep quality."

Once you've reviewed that, add the next angle: "Now add family and parenting desks, for the angle on how the screen-time finding affects parents."

Then: "Now add consumer-tech writers, for the screen-time habits angle."

2) Name all the angles at once (faster). If you'd rather build the whole list in one go, describe the campaign and list the angles in a single request.

Example: "Build one list for this sleep study. Cover three angles: the main health finding for wellbeing reporters that screen time before bed lowers sleep quality, then the parenting angle for family desks, and the screen-time angle for tech writers. Keep them all in one list."

Both approaches will build a single list. Adding angles one at a time just gives you more control as you go.

Avoid: "Build a national list of wellbeing reporters for the sleep study, then a Missouri-local list of data reporters, then a list for a separate diet study."

That's three separate lists.

Build them one at a time instead: finish and refine the first, then start the next as a new list, or in a new conversation.

If the assistant stops partway.

When you add angles one at a time, the assistant may pause between them.

But if you asked it to cover several angles at once and it stops after the first, simply ask it to continue: "Run the other angles into the same list."

You rarely nail the perfect list on the first pass, and that's fine. The assistant is built to be conversed with: react to what it returns and tell it why something is off.

Vague dissatisfaction ("these aren't great") gives it nothing to work with; specific feedback does.

Weak: "These aren't quite right, try again."

Strong: "Too many general lifestyle reporters here. Drop anyone who only covers product round-ups, and weight harder toward reporters who've written about peer-reviewed health research in the last year."

This refinement loop is where the tool earns its keep - the first list is a starting point, not the answer.

If you need 30 names, say 30. If you only care about journalists active in the last 6 months, say so. Defaults are sensible but rarely match exactly what you need, and stating it saves a refinement round.

Weak: "Give me a good-sized list of recent journalists."

Strong: "I need around 25 journalists who've published relevant coverage within the last 12 months. Quality over quantity - I'd rather have 20 strong matches than 40 loose ones."

A great brief = who you're pitching for + the actual story angle + the kind of journalist you want + your hard constraints. Say the obvious things out loud, lead with the story rather than the topic, and treat the first list as a draft you refine - not a verdict.