"Where should we advertise?" is usually answered from habit. B2C goes to Meta, B2B goes to LinkedIn, everyone runs search. The habit is often right, but you can check it in ten minutes against the one source that does not have an opinion: the public ad libraries, which show what your competitors are actually paying for, this week.
Here is the check, then the trap.
The ten-minute check
Pick five competitors. For each one, do three lookups.
Meta. Go to the Ad Library, choose all ad categories, and search the brand name. Open the page that is actually theirs (watch for resellers and fan pages using the name). Note three things: whether the page has any active ads, how many, and the oldest start date in the list. A page with forty active ads and a six-month-old one is buying Meta seriously. A page with two ads from last week is testing.
Google. Go to the Ads Transparency Center and search the advertiser. If they are verified you will see their ads by format with first-shown dates. Note whether they run search, display, video or shopping, and roughly how many. No listing at all means either they are not advertising on Google or they are advertising under a different verified name, so search the parent company too before you conclude anything.
LinkedIn. Go to the LinkedIn Ad Library and search the company. You will see current and recent ads with the company page attached. Note whether there are any, and what kind: single image, document, video, thought-leader posts from a named person. There are no dates or spend to read, so the count and the formats are the signal.
Put the fifteen answers in a grid: competitors down the side, platforms across the top, a count in each cell. Now you have the market's own answer to "where do we play," and it will sometimes surprise you.
What the grid tells you
An empty column is a finding. If none of your five competitors runs LinkedIn ads, either the audience is not there or nobody has tried. Both are worth knowing before you spend, and they lead to opposite decisions.
A full column with old dates is a crowded room. Everyone is buying Meta and has been for months. You can still win there, but you will win on the angle, not on showing up.
A single advertiser in a column is a hypothesis to watch. One competitor quietly running Google Shopping while the rest ignore it means someone has found something, or someone is wasting money. Check again in a month.
Absence is not proof. A brand might be advertising in a country the library search is not showing you, or under a different page. Treat an empty cell as "not observed on this date," and write the date down, so the next check can tell you whether anything changed.
The trap: asking a model instead of looking
It is tempting to skip the lookups and ask an AI assistant which platforms your competitors use. We measured what happens when you do. In a published benchmark we gave the same market questions to a frontier model with no data and compared its answers with the ad libraries. It got the platform verdicts right. It also named advertisers for each platform, and 43% of the ones it named were not observed advertising on the platform it claimed. Its top theme for Meta was absent from the real market. Plausible, confident, wrong.
The same model with real library access got the composition right. The lesson is not that models are bad at this. It is that the answer lives in the libraries, and anything that has not read them is guessing in a convincing voice. The full benchmark, including where the model matched us, is published.
Doing it every week
The ten-minute check is fine once. The useful version is the same grid, updated every week, with dates, so you can see a competitor arrive on a platform or leave it. That is the part that stops happening by hand.
MessCube keeps the grid for you. Tell it who your audience is and it finds the brands talking to the same people, reads their ads on Meta, Google and LinkedIn, and shows who is reaching them on each platform. When a platform turns out thin, it says so and shows how many ads it read to find out. The market read is $39 a month and you can start free with a preview of your own market first.