There are four research reports: competitors, market trends, audience, and reviews. Each costs 20 credits and takes ten to twenty minutes. You can close the tab; they keep running.
The thing worth understanding is what happens after they finish, because it is not what most people expect.
The reports are inputs, not outputs
A research report is a document you can read. Fine. But its real job is to sit in the campaign's context and get read by everything you generate afterwards.
Generate Google Ads headlines before running competitor analysis and you get headlines about your product. Generate them after, and you get headlines that lean on the gap your competitors leave open, because the generation can see what those competitors claim.
Same for the others. Audience research changes who the copy is talking to and which channel it assumes. Review research changes the vocabulary, because it hands over the actual words customers use. Market trends changes what gets treated as urgent.
This is why running research after you have generated everything is the expensive mistake. You paid for the insight and then did not use it.
What order to run them in
If you run one, make it competitor analysis or reviews. Competitors sharpen your positioning; reviews sharpen your language. Both are immediately visible in the next thing you generate.
If you run two, take one of each of those.
Audience research earns its place when you are entering a market you do not know, or when you suspect you are writing to the wrong person. It is a deep pass across real profiles, threads, store reviews and community chatter rather than a demographic template.
Market trends is the one to run when timing is part of the pitch, or when a board deck is in your near future.
Everything is verified before you see it
Deep research reports go through a verification pass: every link is opened, every quote is checked. You will see a line summarising it, along the lines of how many claims were researched, verified, corrected or replaced.
Anything that could not be confirmed is marked rather than quietly kept. If too little of a report survives verification, the report is not delivered at all and your 20 credits come back.
That said, the accuracy notice on every report is there in earnest. Research is AI-generated, it is mostly right, and models can still be confidently wrong about things nobody wrote down. Check anything you would hate to be wrong about in public.
The pre-check on reviews
Review research runs a quick check first, before charging you, to confirm real public reviews of your exact product exist. No footprint, no charge. You then get the option to paste your own links (a G2 page, a Trustpilot profile, a Reddit thread) and have those analysed instead.
This exists because "we found no reviews" is a genuinely common answer for younger products, and being charged 20 credits to learn it would be irritating.
Frequently asked questions
Do I need to re-run research for each new campaign?
Yes. Research belongs to the campaign it was run in, because it was scoped to that brief. A campaign for a different product needs its own.
How long do the reports take?
Ten to twenty minutes is typical. They run in the background, so you can close the tab, leave the page or start something else. Verification adds another one to three minutes at the end.
Can I run the same report twice?
You can, and it costs another 20 credits. Newer reports can be marked as the latest, which is the one that feeds downstream generations.