How is AI Research metered, and what do paid sources cost?

"AI Research" covers the basics of what a run costs; this article goes deeper on how the metering actually works, source by source. For the full prices of every feature, see "Price list: what every action costs."

Metered by actual effort, not a flat fee#

Every reasoning step the AI takes (deciding what to search, weighing a result, checking a match) is billed by the amount of text it processes — the usual AI costs, same as an AI command or an AI column elsewhere in ListPlus. Starting a run needs a small minimum balance, and you also set a spending budget for it (see below); there's no fixed price per match or per run.

Every source is its own switch, with its own price#

Under Allowed AI tools (step 3 of the form) each source is a separate checkbox in a folding group (Find, Signals, Community & intent, Verify & enrich), and each has its own price and unit. A checkbox allows a source; it doesn't start a query — the agent calls a paid source only when the brief needs it, so a ticked source the brief never needs costs nothing:

A call is billed whether or not it turns up something useful, since the underlying source bills for the call itself; an empty database page, or one repeating the exact same search within 30 days, is free. Reading LinkedIn posts is two separate charges: the search itself, then a further charge only for the posts whose people the agent actually reads.

The budget you set#

Credit limit (under Budget & model) sets the credit budget for the run — you pick it from a list before you start (the options are in "Price list: what every action costs"), covering the AI's own reasoning together with every paid source call. The run paces itself to stay inside that ceiling and stops itself once it's spent, keeping everything found so far.

Keep searching until … are found (on by default when you set a target) starts another round when the AI ends one early with fewer matches than you asked for — up to 5 times, never past the credit limit. Model decides what each step costs: the default model is fast and cheap, the stronger one has the best judgment but costs many times more per step, so the same credit limit buys far fewer steps with it.

Watching the cost live, and stopping anytime#

While a run is active, a live counter next to your match count shows credits used so far and updates continuously. Click Stop in the header at any point to end the run immediately — everything found up to that moment stays, but credits already used aren't refunded.

Resuming costs again, and running out of credits#

Clicking Resume — also on a finished run — opens another pass the same size as what you originally chose (matches and budget both), metered exactly like the first pass: reasoning steps plus any paid-source calls. If you run out of credits mid-run, the run stops itself, keeps what it found, and the credits it used aren't refunded; top up and click Resume to continue (details: "Not enough credits").

Related topics#

"AI Research" · "Price list: what every action costs" · "How do I write a good AI Research brief?" · "Getting AI Research hits into a list automatically" · "Not enough credits".

FAQ#

What am I actually paying for during an AI Research run?#

The usual AI costs — the agent's own reasoning as it reads, searches and verifies, billed by usage — plus a per-call charge for any source you've switched on that isn't free. There's no flat fee per match or per run.

Which sources are free and which cost credits?#

Web search, reading web pages, job postings, news, Hacker News and finding a real LinkedIn profile are free. Database lookup, LinkedIn people, LinkedIn posts, Reddit and X all cost credits per call — the table above shows which; prices are in "Price list: what every action costs"; the tech-stack lookup is currently not available.

Are paid sources like Reddit or X charged once per run?#

Per call — the agent can call an unlocked source several times in one run, and each call is charged even if it finds nothing. Reddit is charged per batch of 25 posts or comments requested, and X per batch of 100 posts requested.

What is the spending budget ("at most"), and can I change it mid-run?#

Credit limit, under Budget & model, sets the credit budget for the run — you pick it from a list before you start (the options are in "Price list: what every action costs") — covering both the agent's reasoning and every paid source it calls; the run stops itself once it's spent. You set it before starting, and a resumed pass reuses the same ceiling on top of what's already spent.

Where do I see how many credits a run has used?#

A live counter next to your match count updates continuously while the run is active, showing the running total spent so far.

Does clicking Resume on a finished research cost credits?#

Yes — each Resume opens a new pass the same size as what you originally chose and is billed the same way: reasoning steps by usage plus any paid sources it calls, up to the budget you set again.

Can I stop a research before it finishes?#

Yes — click Stop in the header at any time. The run ends immediately and keeps every match already found, and credits already used aren't refunded.

What happens if my balance runs out mid-run?#

The run stops itself and keeps everything found so far; the credits already used aren't refunded. Top up your balance and click Resume to continue with a fresh pass.