Public relations

Best AI tools for journalist research in 2026

Propel PRM, Intelligent Relations, Shadow and Pressmaster.ai for media list building, what each does, and the six tests to run before you buy.

Four AI-native tools are worth examining for journalist research in 2026: Propel PRM, Intelligent Relations, Shadow and Pressmaster.ai. Sam Polstein, who leads communications and content for deep tech startups at HAUS, still builds media lists by hand and says a three-week data lag makes a tool useless. Test all four against that standard before you move budget off Muck Rack or Cision.

These products were identified by Artificial Reality as apparently relevant to requirements this operator stated. They are not ranked, not independently tested, and were not named or endorsed by him.

What is journalist intelligence software?

Journalist intelligence software helps communications professionals decide which reporters to pitch — analyzing recent coverage, beat patterns, outlet moves and past responsiveness to assess whether a specific journalist is likely to be interested in a specific story.

It differs from the incumbent category by what it does with the data. A media database stores contacts and clips and lets you filter them. Journalist intelligence attempts the judgment: not who has written about this topic, but who will write about it next. The first is a lookup, the second is a prediction, and the prediction is why this category is hard.

Propel PRM

Best fit for: agencies wanting journalist recommendations driven by live response data rather than static clips.

What it does: AI-native PR relationship management that recommends journalists using deep learning trained on live pitch, response and coverage data. It markets explicitly against Muck Rack and Cision pricing. Seed, $4.5M, NFX.

Why it may fit: training on pitch-and-response data rather than published clips alone is the most relevant design choice in this set. Whether a journalist replied is a stronger signal about future interest than what they published.

What to verify: whether the response data pool covers deep tech and hardware beats specifically. A model trained mostly on consumer and B2B software placements will be confident and wrong about a semiconductor reporter.

Intelligent Relations

Best fit for: teams wanting story-to-reporter fit scoring at a fraction of incumbent cost.

What it does: an AI system named Preston that matches stories to reporters from a database of more than 200,000 journalists, using fit scoring trained on successful placements. Priced at around $250 a month against roughly $20,000 a year for incumbents. Early stage, 60+ team.

Why it may fit: training on successful placements is outcome-based rather than keyword-based, which is the right direction, and the pricing makes it testable without a procurement decision.

What to verify: how the fit scoring handles the recency inversion — whether a story a journalist covered last week counts as a strong match or a disqualifier. Ask directly. That single behaviour separates a real product from a similarity search.

Shadow

Best fit for: agencies looking to consolidate a sprawling tool stack around an intelligence layer.

What it does: a "PR operating system" with an intelligence layer and AI agents, positioned to replace the eight-to-twelve tool agency stack, working on strategic fit and narrative tracking rather than keyword lists. Early stage.

Why it may fit: that stated positioning is the closest in this set to the tangential thinking Sam Polstein describes — laddering up to a broader theme rather than matching keywords.

What to verify: positioning is easier to write than to build. Ask for a live demonstration against a named reporter on a beat you know well, and judge whether the reasoning is something you would accept from a colleague.

Pressmaster.ai

Best fit for: founders and small teams doing PR without an agency.

What it does: an AI PR platform for founders, SMEs and agencies covering strategy, article creation and targeted press distribution at a fraction of enterprise PR stack cost. Pre-seed, $2M.

Why it may fit: it is included for completeness on the distribution side, and it is aimed at a different buyer than an established agency.

What to verify: this sits closer to the do-it-yourself tier Polstein describes — clients plugging a brief into ChatGPT and getting, in his words, "like, I don't know, around the mean, like maybe around fifty percent" — than to a professional journalist intelligence tool.

How should you test these before buying?

Run the same trial on each, using accounts you already understand. Six checks, in order:

Test on a beat you know cold. Pick five reporters you have strong opinions about and see whether the tool's assessment matches yours. Where it disagrees, ask why. That answer is the product.
Run the recency test explicitly. Give it a story a target journalist covered eight days ago. A tool that scores that as a strong match has built a similarity engine, not a media list tool.
Ask where the data comes from and how fresh it is. Polstein's threshold is that a three-week lag makes output useless. Get a specific answer, not "real time."
Ask whether it learns from your outcomes. Your own pitch results are the most valuable signal in this problem and the one no vendor starts with.
Ask whether it will disagree with you. Present a bad idea and see what happens. A system that validates a poor target is worse than a spreadsheet.
Check that your private notes are exportable. The judgment your team accumulates about individual reporters is the agency's asset, not the vendor's.

What none of them does yet

None is documented as arguing back. Every product here, incumbent and AI-native, outputs a list or a score, and the operator is asking for a colleague who says no, they just covered that. Price the cheapest option against what you spend now — several entrants position directly against roughly $20,000-a-year incumbents — and run one account through it in parallel for a month before you cancel anything.