ads.txt checker

ads.txt is one half of a pair. You publish it to declare which companies may sell inventory on your domain. The ad platforms publish sellers.json to declare which publishers they represent. A buyer reads both, follows the supply chain attached to the bid, and confirms that the path from your domain to the seller offering the impression is one both ends agree on. Where the two do not agree, the bid is dropped.

This is why the states are asymmetric in a way that surprises people. No file at all leaves your inventory unverified, which most buyers still tolerate. A file that exists but omits a partner you genuinely work with actively blocks that partner, because you have now made a positive declaration and they are not in it. The second state is the one that costs money, and it is usually reached by adding a second network and forgetting the file.

It is checked in the technical audit for the same reason as robots.txt and the sitemap: a plain text file at a fixed path that any machine can read and that quietly stops working when nobody is watching it. Search engines do not read it and it has no bearing on rankings.

Once you are serving ads, the failure announces itself in your AdSense account rather than on your site. The wording is an earnings-at-risk or unauthorized-inventory notice, and it means Google fetched your file and did not find a line naming itself with your publisher ID. Pages keep loading, the ads simply stop filling.

Changes are not instant. Crawlers re-fetch the file on their own schedule, so allow a day or two after an edit before deciding it did not work, and confirm the new file is actually being served rather than a cached copy from a CDN that still has the old one.

Keep it current when your setup changes. Adding a network means adding its line, leaving one means removing it, and working through an ad management partner usually means their lines rather than yours, with your own DIRECT line kept alongside. An old line for a company you no longer work with is not dangerous, it is just an authorisation you have left standing for no reason.

Here is a four-line file of the kind this check fails every day. Line one reads google.com, pub-4739284712047392, DIRECT, f08c47fec0942fa0 and is a valid record. Line two is the same record pasted from a word processor, where the quotes turned curly and the commas became a lookalike character from another alphabet. Line three reads partner-network.com pub-88231 RESELLER, with the commas simply missing. Line four is a comment starting with # that explains line three. In a browser all four look equally tidy. To a parser, line one is the only record that exists, so the partner on line three is not authorized and does not know it.

The way to see what a machine sees is to stop using the browser. Fetch the file with curl and read the raw response: the status code, the Content-Type header, and the first bytes of the body. A surprising number of sites return their 404 page at that path with a 200 status, or serve the file wrapped in the site template because the CMS treated /ads.txt as a pretty URL. Both count as no file, and neither is visible from the editor where the content was typed.

The organisational version of the failure is quieter. A site starts with AdSense, one line, correct. A year later it adds a video player that brings its own advertising demand, then joins an ad management service. Each of those partners publishes the exact lines they need, usually a dozen or more, and the moment they go live without those lines in your file, their auctions are the ones being dropped. The file was right on the day it was written and wrong by accretion, which is why it belongs on the same review-after-every-change list as robots.txt.

Two placement details finish the job. Subdomains that serve their own ads need their own file, because a buyer checking blog.example.com does not read example.com/ads.txt unless a subdomain declaration points there. And a CDN can cache the old copy long after you deployed the new one, so after any edit, fetch the live URL from outside and confirm you are reading the version you think you shipped, the same habit the technical audit applies to every plain-text file it checks.