Overview
A paid-search dispute is not usefully sorted by the platform it ran on. It is sorted by what it turns on — and that division decides, before anyone opens an export, whether an examination of the ad account is worth what it costs.
Six disputes are covered here, in three groups. Fraud matters turn on the traffic and how it was delivered. Contract matters turn on the agreement, and the account can only show where the money went. Conduct matters turn on what was actually done inside the account, and on whether there was any standard to depart from. The groups are no filing convenience. Each names the fact that has to be established, and that fact decides which record you need and who holds it.
Each of the six pages opens with one of three verdicts: the data settles it, the data plus discovery settles it, or this is not a data question. That third verdict sits on two of the six. I would rather an attorney read it here than pay to find it out.
Start from the fact in dispute, not the label on the claim
The caption on a pleading tells you very little about which record answers it. A claim styled as fraud frequently turns on what an agreement permitted. A breach claim frequently turns on whether a tag was firing.
So the first question is not what is pleaded. It is: what single fact, if established, moves this, and is that fact recorded anywhere? Sorting by the answer to that is what produces the three groups.
Fraud: the dispute is about the traffic, and the record is split
Click fraud and invalid traffic, and affiliate and lead-generation fraud, share a structure: the fact in dispute is whether traffic that was billed was what it was represented to be, and the record is split down the middle. The platform holds one half, the advertiser holds the other, and neither half is sufficient alone.
The platform's half is thinner than most attorneys expect. An advertiser gets a count of filtered clicks and a credit against the bill. There is no click-level ledger tying a filtered click to an address, a device or a source, and no platform publishes why any particular click was ruled invalid. The reason for the silence was stated on the record by a court-appointed expert as early as 2006 and has not changed: publishing the operational definition of an invalid click hands it to the people building the traffic. The consequence is that the platform's number and a detection vendor's number will not agree, and the gap is not itself evidence of anything.
The advertiser's half — server logs, tag data, the click record on the merchant's own side — is the only part controlled end to end. It is also, routinely, the part with the shortest retention. Both fraud matters carry a "data plus discovery" verdict for that reason: an expert moves them, but not from the ad account alone.
Contract: the agreement decides it, and the account audits the money
Agency and client disputes, and franchise and cooperative advertising disputes, are contract matters in advertising costume. What the agreement required, what the fund's terms permitted, what the scope of work covered — none of that is in an ad platform, and no volume of account data supplies it.
What the account does supply is narrow and genuinely useful. Ownership of an advertising account is a status inside the platform, held by exactly one party, and it is not conferred by having paid for the media. Platform cost is a recorded figure and can be set beside an invoiced figure. The date a conversion tag stopped firing is recorded, and so is the condition an account was handed back in. That is an audit of where the money went, which is often the difference between a settlement number and a guess.
What it does not supply: whether the client approved a change, whether a markup was disclosed or permitted, whether a charge against a shared advertising fund was within the fund's terms. Those live in email, ledgers, media plans and the agreement itself, and they arrive through discovery or not at all. One of the two contract pages carries a "not a data question" verdict outright, and I would rather say so than sell an examination that cannot reach the fight.
Conduct: what was done, against a standard that does not exist
The two conduct matters are the least alike of any pair on the site.
Competitor and trademark bidding is, in the reported decisions, largely a question about what appeared in the advertisement rather than what was targeted in the account. That is a record question: ad text, assets, display URLs, landing pages and the change history behind them are all recorded, and the distinction between the two facts can be documented cleanly. What no export reaches is whether consumers were confused. That is survey and fact-finder territory, and an expert offering it from account data is claiming more than the record carries.
The standard-of-care page is the other end of the scale, and it exists because of an absence. There is no published, citable standard of care for paid-search management — no licensing board, no accrediting body, no professional association issuing practice standards with disciplinary authority behind them. That is a finding, stated as of the research date of August 14, 2026, with no contrary source located. It is no gap being papered over.
The practical effect is direct. An opinion framed as "no competent manager would have done this" rests on a document that cannot be produced, and asking for its issuer and its date is a fair question to put to anyone who offers it. What can be established is what was done, what it cost, and what the platform's own dated documentation said about it at the time. Certifications never fill the gap: a platform certification is a 75-minute exam requiring 80 percent to pass, valid for one year, and a partner badge is a commercial threshold keyed to an optimization score and a 90-day spend requirement. Neither is a standard of care, and neither establishes competence or its absence.
What retaining an expert will not do
It will not produce an industry standard for paid-search management, because there is not one to produce. It cannot tell you who was physically at the keyboard — platform logs record the identity a change was made under, not the person operating it. It cannot establish what a competent manager would have achieved, which is a counterfactual and belongs in a model, argued as a model. And it cannot recover a record whose window has already closed; the clocks on this evidence run from today, not from the date of the conduct, which is why the timing question comes before the merits question on all six.
Whether any of it is actionable is a question for counsel; I am not an attorney. What I can tell you is which of the six disputes the account record moves, how far it moves them, and where it stops.
Published