How to Find Competitor Ad Spend: 8 Sources and Why Every Number Differs

Start with the uncomfortable part. There is no legitimate public source that reports what a commercial advertiser spends. Platform transparency tools were built to disclose who is advertising and what they are saying, not how much they are paying, and the spend figures they do publish are limited to political and social-issue advertising.

That sounds like bad news and mostly is not, because the things you can obtain exactly turn out to be more decision-useful than a budget number would be. You can see every creative they are running, in which markets, since when. You can measure their bidding pressure on your own keywords. And for a listed competitor you can bound the total from audited filings.

1. 1Replace the question with one that has an answer. "What do they spend?" has no reliable public answer. "Which channels are they in, with what message, at what intensity, and are they outbidding us?" has four, and each is measurable.
2. 2Pull every live creative from the ad libraries. Search the competitor in the Meta, Google and professional-network ad libraries. You get the creative, the format, the regions and the start date at no cost, which is the exact part paid tools charge to summarise.
3. 3Read cadence rather than volume. Count how many distinct creatives are running and how often new ones appear. A steady stream of new variants indicates active testing and a real budget; three ads unchanged for six months indicates an always-on placeholder.
4. 4Measure the pressure they put on you. Open Auction Insights in your own ad account. Impression share, overlap rate and position-above rate tell you exactly how hard a named competitor is bidding against you, and this is a measurement rather than an estimate.
5. 5Set the ceiling from their filings. If they are listed, take total sales and marketing expense from the latest filing. Advertising is a fraction of it, so the figure is an upper bound rather than a spend number, and it instantly disqualifies any estimate that exceeds it.
6. 6Use estimators last, as a tiebreak. Third-party spend figures are modelled traffic multiplied by modelled cost per click, so two tools will disagree by multiples. They are useful for direction and relative movement over time, and misleading as absolute numbers.
7. 7Write the conclusion as a range with its basis. State the band, name the method, and date it. "Between roughly X and Y, based on filed marketing expense and observed auction overlap, as of this month" survives scrutiny; a single confident number does not.

The ad libraries exist largely because regulation requires them. Under the EU’s Digital Services Act, very large platforms must maintain a public, searchable repository of the ads they carry, covering the period an ad runs and for a year afterwards. The obligation is about transparency of who is advertising, which is why the disclosure stops short of commercial spend.

1. Meta Ad Library

Search the competitor’s page name and filter by country and by whether the ad is still active. You get the creative, the format, the platforms it runs on and the date it started. Sort by start date and note how many distinct creatives launched in the last month: that count is your best proxy for the size of the programme, because testing volume scales with budget.

2. Google Ads Transparency Center

Search by advertiser and you get their verified identity, their location, and the creatives they have run across Search, YouTube and Display for roughly the last 30 days. There is no spend and no impression data for commercial advertisers, so read it for message and channel mix rather than budget. The verified-advertiser identity is itself useful when a competitor runs ads through an agency or a differently named entity.

3. LinkedIn Ad Library

Search the advertiser by name to see their active sponsored content, with filters for keyword, country and date. For a B2B competitor this is usually the most representative view of their paid programme, since it is where the audience they actually want sits. No spend or impression figures are published.

4. Other DSA ad repositories

TikTok, X, Microsoft and other very large platforms maintain equivalent repositories under the same rules. Check them when a competitor’s audience plausibly sits there. Coverage and usability vary considerably between platforms.

5. Your own Auction Insights report

This is the strongest source on the page and it is inside your own ad account. Auction Insights lists every advertiser competing in the same auctions as you and reports impression share, overlap rate, position-above rate and top-of-page rate against each. It is measured rather than modelled, it names competitors directly, and it is scoped to the keywords you actually care about. A competitor whose impression share climbs 15 points over a quarter has increased their budget on your terms, and you know it as fact.

6. Their filings and annual report

For a listed competitor, sales and marketing expense is audited and published quarterly. Advertising is only a portion of it, alongside salaries, events, content and tooling, so treat it strictly as a ceiling. Its value is disqualifying nonsense: if a tool estimates annual ad spend above their entire filed marketing expense, you can discard the estimate without further argument.

7. Third-party spend estimators

Semrush, Similarweb, Ahrefs and SpyFu publish estimated paid spend. The method is fundamentally the same everywhere: model the traffic, model the cost per click, multiply. Two estimates multiplied together compound their error, which is why two tools routinely disagree by multiples for the same company. They are genuinely useful for tracking one advertiser’s relative movement over time, and misleading whenever quoted as an absolute number.

8. Sponsorships, rate cards and agency case studies

Newsletter sponsorships, podcast reads, conference booths and review-site placements often have published rate cards, so a visible placement converts directly into a known cost. Agency case studies sometimes disclose budgets outright. This is the only category where you can occasionally attach a real number to a real placement.

1. 1Bound it before you believe it. Check any estimate against filed marketing expense where one exists. Anything above that ceiling is wrong, and a surprising number of published estimates are.
2. 2Cross-check two estimators. If they disagree by more than a factor of two, and they often will, you have a range rather than a number. Report the range.
3. 3Confirm the channel mix from the libraries. An estimator that models only paid search will badly understate a competitor running mostly paid social, and it has no way to tell you that it is doing so.
4. 4Prefer the measured signal where you have one. Impression share from your own account beats every external estimate for the keywords that matter to you, because it is an observation rather than a model.

The budget question is the one people ask, and it is usually not the one worth answering. An ad that has been running for six months has survived continuous optimisation against real money, which makes it the most rigorously tested statement of positioning a competitor produces. It is free to read and it is more reliable than their homepage copy.

For each recurring creative, record four things:

• The promise in the first line, which is the benefit they have decided leads.
• The segment or role named, which tells you who they are prioritising this quarter.
• The proof offered: a number, a customer, an analyst mention, a guarantee.
• The objection pre-empted, which reveals what their buyers push back on.

Feed that into a messaging framework or a positioning map, and the ad library stops being a spend estimator and becomes the cheapest positioning research available.

Advertising is public by definition, and the ad repositories exist so that anyone can inspect them. Reading them is the intended use, not a grey area. Three limits apply.

• Do not copy their creative. Copy, imagery and layout are protected works. Learning from a competitor’s message is normal practice; reproducing their asset is infringement.
• Be careful bidding on their brand. Bidding on a competitor’s brand name as a keyword is permitted on the major platforms in most markets, but using their trademark inside your ad text is restricted and can breach both platform policy and trademark law. Rules differ by country.
• Do not click a competitor’s ads to drain their budget. It is click fraud, platforms detect it, and it exposes your company rather than theirs.

• Their actual budget. Known only inside their ad accounts. Proxy: filed marketing expense as a ceiling, creative volume as an intensity signal, and impression share as relative pressure.
• Their cost per acquisition or ROAS. Requires their conversion data. Proxy: how long a campaign keeps running. Ads that persist for months are almost certainly working.
• Their targeting parameters. Disclosed only in aggregate, and mainly for EU audiences under the transparency rules. Proxy: the segment named in the creative itself, which is usually explicit.
• Ads inside AI assistants. Ads served in ChatGPT appear inside private one-to-one conversations, and OpenAI publishes transparency reporting rather than a searchable per-advertiser archive, so there is no ad library to check. Proxy: run the buying-intent prompts your customers would use and record what appears, and watch competitors’ job ads and agency announcements for the channel being staffed.

Check the ad libraries monthly, and immediately after a competitor launches a product or you lose several deals on positioning. Pull Auction Insights on the same schedule as your own reporting so the two line up. What you are watching for is a new creative theme, a new channel or a step change in impression share, not a fluctuation in an estimated figure.

Record the change and what it makes you do, in a competitive landscape or your regular competitive update. Flares watches ad and messaging changes continuously, which matters here because a campaign launch is time-sensitive in a way that a pricing page rarely is.