What an SEO Expert Witness Does

A narrow role, deliberately

An SEO expert witness is retained to answer a technical question in a way a decision-maker with no technical background can follow and check. That framing sounds modest and it is the whole discipline. Most of what goes wrong in this work is somebody exceeding it.

The question is almost never “was this good SEO.” Quality is a matter of professional judgment and it makes a poor subject for an opinion that has to survive challenge. The questions that can actually be answered are factual: was the work described in this agreement performed; did this party's conduct cause the decline being claimed; was the ranking that was promised achievable at all; was a competitor's brand term used to divert search traffic, and can that be shown from records rather than assumed from an outcome.

The work is turning material that lives in server logs, analytics exports, archived pages, registrar records, crawl data and search console accounts into findings a non-technical reader can weigh — with the uncertainty left in rather than sanded off. Search evidence is circumstantial by its nature. An analysis that presents it with the confidence of an accounting record will not survive competent challenge, and it should not.

The disputes that produce this work

Matters involving search evidence tend to fall into recurring shapes, and recognizing the shape early tells you what to preserve.

Performance and contract disputes — whether an agency or contractor delivered what an agreement described, and whether the deliverables produced match the deliverables invoiced.
Claimed interference — allegations that someone manipulated links, content, listings or crawl access to damage another party's visibility.
Trademark use in search — a brand term appearing in paid search, anchor text, page copy, metadata or a competitor's domain name.
Asset valuation — transactions where search visibility formed part of what was bought, and the buyer says it was not what it appeared to be.
Post-migration collapse — a rebuild, replatform or domain change followed by a traffic loss, and a disagreement about whose decision caused it.
Reseller, franchise and affiliate conduct — branded bidding, duplicated sites, and channel behavior that violates an agreement.

One thread runs through all of them: somebody is asserting a causal story about search, and the records either support it or they do not.

How an engagement actually runs

The sequence is fairly consistent, and knowing it helps counsel budget both money and calendar time.

Conflicts first. Parties, related entities, counsel and any prior involvement are checked before I look at anything substantive. This takes a day and it is not a formality.

Scope and role in writing. What question is being asked, what materials will be available, and whether the engagement is advisory or intended to be disclosed. Those are counsel's decisions under the rules of their jurisdiction. They should be made deliberately rather than discovered later.

A preliminary assessment. Before any substantial analysis, a short review of what exists tells you whether the question is answerable at all. If the data needed to answer it was never collected or has already been destroyed, that is worth knowing in week one rather than month four.

Discovery input. Frequently the highest-value contribution, and it happens early or not at all.

Analysis, then a draft. The analysis is built to be re-run. The draft states the question, the materials, the method, the findings, the assumptions and the limits.

Response to the other side's analysis where there is one, which is a different exercise from producing your own — it is about method, sourcing and whether the stated conclusions follow from the stated evidence.

What to demand while it still exists

Search evidence expires on its own schedule, and preservation letters written before anyone knows what to preserve reliably miss the things that matter. The list worth asking for by name:

Raw server access logs rather than summary reports, covering the periods at issue. These are frequently rotated within weeks and are often held by a hosting provider or content delivery network rather than by the party.
Direct analytics account access rather than exported reports, including the configuration and filter history, since filters are a common source of numbers nobody can reproduce.
The Search Console property export for every property variant, noting that this data ages out on a rolling basis and cannot be recovered afterward.
Content management revision history, which establishes what a page said and when it changed from the site's own system rather than from an archive.
Tag manager container versions, which date tracking changes precisely and frequently explain apparent traffic movements.
Registrar and DNS records, including account login history and authorization code requests where a domain is involved.
The agency's contract, deliverables, work product and monthly reporting, plus the communications around them.
Any rank tracking account that was running at the time, because if nobody was tracking the terms on the dates that matter, no tool can reconstruct them later.

That last point deserves emphasis. There is no authoritative historical record of search rankings. Tools store samples taken from chosen locations and devices on the days they happened to run. An expert who offers to reconstruct past positions without a contemporaneous record is describing a model, not a measurement.

Building an analysis that holds

The method matters more than the conclusion, because the method is what will be examined.

Preserve before analyzing: work from copies, leave originals untouched, and record a cryptographic hash of each file at collection so the version relied on can be identified beyond argument. Make every number traceable to its export, account, property, date range, filter set, segment and time zone — “traffic fell sixty percent” is not a finding until a reader can reproduce it. Capture web pages as responses, with status code, headers and raw HTML, alongside a full-page rendering showing the URL and the system clock; a picture of a browser window is the weakest form of web evidence there is. Prefer scripts to manual steps, so the analysis can be re-run by somebody who does not trust you, which is the point.

On causation specifically, the discipline is mechanism testing. Every alleged cause leaves a distinct fingerprint. If the claim is deindexing, impressions should fall and not merely sessions. If the claim is a tracking failure, sessions fall while orders recorded in the accounting system hold. If the claim is a manual penalty, there is a message in the property's own console account. If the claim is a link attack, the acquisition should be visible in the link data and should precede the loss rather than follow it. Then segment — branded against non-branded, page group, device, country — and find a control. A decline that appears everywhere at once, including in areas the alleged conduct never touched, is telling you something other than what the complaint says.

How to evaluate an expert

If you are retaining someone in this field, these are the things I would check, and they are the same things I would examine in an opposing analysis.

Will they tell you what the evidence cannot show? An expert who has never written the sentence “the available data does not support that” has not been doing the work honestly.
Is the method reproducible? Ask how a finding would be re-derived by somebody else. If the answer involves a proprietary tool and a judgment, that is a problem.
Do they distinguish observation from inference? That a page existed on a date is an observation. That a ranking change caused a revenue change is an inference, and it needs labeling as one.
Do they stay inside the technical question? Willfulness, bad faith and unfair competition are legal characterizations, not expert findings.
Is there a public record of their thinking? Published writing and conference material over a period show whether views are consistent and whether they have been stated in advance of any engagement.
Are they independent of the work at issue? The incumbent agency is a fact witness about its own performance, which is a difficult position to be useful from.

Where I fit, and where I do not

I have been practicing SEO since 1996 and the technical work in a disputed matter is the same technical work I do outside one — log analysis, crawl and index reconstruction, link profile examination, registrar and DNS record reading, analytics reconciliation. What changes in a litigation context is the standard of documentation and the discipline about what may be claimed.

What I will not do: adopt a conclusion before the analysis, express a level of confidence the records do not carry, opine on the legal question, or produce a damages figure built on assumed conversion rates and assumed margins that nobody inside the business would recognize. I also will not discuss engagements. This work is confidential as a matter of course, which is why this site describes the practice and names no matters, no clients and no outcomes.

Engagements are arranged through counsel and handled through Hartzer Consulting. This site is my professional record rather than a place to retain anyone.

Frequently asked questions

What can an SEO expert prove about past rankings?

Less than most people expect. There is no authoritative historical record of search rankings; tools store samples taken from a chosen location, device and query set on the days they happened to run. If nobody tracked the term on the dates at issue, it cannot be reconstructed afterward. What can be established from records is that a page existed with particular content on a date, that a site's own recorded impressions moved between two dates, and that crawl access changed at a known time.

Can you show that a competitor caused our traffic to drop?

Sometimes, and only through mechanism testing rather than timing. Organic traffic moves for many reasons simultaneously, so the analysis has to identify the specific fingerprint the alleged conduct would leave and then look for it: impressions falling rather than only sessions, the link acquisition preceding the loss rather than following it, the affected segments matching where the conduct occurred. Where several sufficient causes overlap and the data cannot apportion between them, that is the finding.

What should we preserve before bringing in a search expert?

Raw server access logs, which are often rotated within weeks and held by a hosting provider rather than the party. Direct analytics and Search Console access, since console history ages out and cannot be recovered. Content management revision history and tag manager container versions. Registrar and DNS records including account activity. The agency contract, deliverables and reporting. And any rank tracking account that was running at the time, because that data cannot be recreated later from anywhere.

Do you work for plaintiffs or defendants?

Both, and the method does not change with the side. The value of a technical opinion comes from being indifferent to the outcome, and an expert whose findings always favor whoever retained them stops being useful once the pattern is visible. In practice this means telling a retaining attorney early when the records do not support the theory they hoped to prove, which is almost always cheaper information than discovering it later from somebody else.

How is an expert different from our existing SEO agency?

The agency is a fact witness about its own work and has an interest in how that work is characterized, which makes it a difficult source of independent analysis. An expert applies a documented method to the records and states what they support, including findings unhelpful to the retaining party. The agency's material — contracts, deliverables, reporting, account access — is usually important evidence in these matters, but it is not a substitute for an independent examination of it.

How early should a search expert be involved?

Before discovery requests are drafted, if possible. The most valuable early contribution is usually not analysis but specification: naming exactly which logs, exports, account accesses and revision histories to demand while they still exist. Much of this material is deleted on ordinary schedules measured in weeks, and some of it, such as search console history, cannot be recovered by any means once it has aged out. A preservation letter written without technical input tends to miss precisely those items.