Ads.txt, sellers.json and app-ads.txt supply chain audit

On a mid-size publisher we usually delete 60 to 80 percent of the lines in ads.txt and monthly revenue does not move. The file grew by accretion: every seller rep who asked for a line got one, every reseller of a reseller got one, and no DIRECT declaration has been checked against a signed contract in years. The cost is not fill. The cost is that a buyer running supply path optimisation cannot prove the path they bought was yours, so they stop buying it, and the drop arrives three months later as soft CPMs nobody can attribute to anything. What we reconcile: Three files against each other, and all three against your contracts. Every line in ads.txt is matched to the exchange's sellers.json record: seller_id, seller_type of PUBLISHER, INTERMEDIARY or BOTH, the declared domain, and whether the record carries is_confidential=1. A confidential record sitting behind a DIRECT line is a conversation to have with that exchange, not a line to keep. DIRECT lines must correspond to an account you are contracted with and paid by. RESELLER lines must trace to a node that will actually appear in the SupplyChain object of a bid request. Where you sit under Google MCM or an equivalent, we declare OWNERDOMAIN, MANAGERDOMAIN and INVENTORYPARTNERDOMAIN per ads.txt 1.1 instead of letting a crawler infer them. On the apps side, each app-ads.txt is checked against the developer URL published in the store listing, bundle by bundle, including the bundles somebody shipped in 2022 and forgot. We also read your own sellers.json, if you filed one. Publishers who resell anything, a network of sites under one entity, a partner's subdomain, a white-label edition, are intermediaries whether or not the record exists, and an undeclared intermediary is the fastest route to being treated as unauthorised supply. What you get back: The corrected files, plus the reasoning line by line as a sheet you can send to a partner who asks why they were removed. The files live in your repository, not ours, with a scheduled job that re-fetches every declared exchange's sellers.json and fails when a seller_id disappears, changes type, or the committed line count drifts. Crawl behaviour matters here: the file is fetched at the root, one off-domain redirect is honoured and no more, and a 404 during a deploy window reads as an empty file. We test for that specifically, because we have watched a five-minute deploy blank a file and cost a weekend of demand. What we will not put in your file: A line because a seller asked for it. A DIRECT declaration for a relationship that is factually a reseller one, however the rep phrases it. A monetisation partner who pays us. We take no rebate, referral fee or revenue share from any exchange, and we will tell you when the honest finding is that you have too few contracts rather than too many lines. Who should not hire us for this: Publishers whose traffic is bought. Sites where most sessions arrive from paid social and leave on the first ad. Anyone who wants the file tidied ahead of a partner audit while the underlying relationships stay undeclared, which is the one version of this work that makes matters worse: a clean file makes a misdeclaration deliberate rather than careless. Crypto and Turkish inventory gets an extra pass. Crypto sites accumulate reseller lines from networks that also carry made-for-advertising supply, and those lines are the first thing a demand partner's own audit finds. Turkish domains often carry legacy lines from local networks that no longer file sellers.json at all. Nobody can validate those, and they come out unless you produce the contract.