Website Traffic Loss Claims

The three things the claim has to establish

An organic traffic loss claim asserts three propositions, and they are not equally hard.

First, that organic search traffic to an identified set of URLs declined. Second, that the decline was caused by identified conduct rather than by anything else. Third, that the lost traffic converted at a rate and a value supporting a damages figure.

The first is usually provable from records. The third is a measurement problem with established approaches. Almost every claim that fails, fails on the second — and it fails there because the party asserting it treated the decline as self-explanatory. A chart that falls is not a causal finding. It is the thing the causal finding has to explain.

This page is about the shape of the claim and the evidence it requires. It stops where damages method begins. How a loss is quantified once causation is established — before-and-after comparison, yardstick selection, seasonality, algorithm updates as a confound, and the division of labor with a financial expert — is a separate body of work with its own reliability problems.

Define the metric before you look at the chart

The word traffic has no fixed meaning across the tools, and cases are lost on this before anyone reaches causation.

Clicks is a Search Console metric: a user clicked through from a Google results page. Impressions counts the times a link to the site appeared in results, whether or not it was seen. Sessions and users are analytics constructs defined by a tag and an attribution model. Pageviews and entrances differ again. None is interchangeable, and Google's documentation explains at length why two of them will never agree.

So the first question is not what happened. It is which number the complaint means, and where it came from. A complaint pleading a 40% decline that means sessions in an analytics property, set against an expert's analysis of clicks in Search Console, is describing two different quantities.

The related discipline is defining the URL set. The site is rarely the right unit. A decline concentrated in one template, one category, or one query class is a different case from a uniform decline, and only the URL-level view distinguishes them. Site totals conceal the pattern that establishes or defeats causation.

The evidence that establishes the decline

Four sources carry this element, with different survival properties.

• Search Console performance data, exported daily for the maximum available range, split by query and by page, with device and country retained. This is Google's own record of how the site performed in Google Search. Its history runs sixteen months on a rolling basis, and the default view of the report shows only the past three months — so a client who pulled Search Console and produced a quarter has produced the default, not the record.
• Analytics data for organic channel traffic. Retention in GA4 is a property setting that has to be checked per property, but standard aggregated reports are not affected by it at all.
• Server access logs, which sit outside any platform's retention policy and record every request, including those JavaScript-based analytics never saw.
• A configured BigQuery bulk export of Search Console data, where one was set up before the incident. Google documents the feature; it helps only if somebody turned it on in advance.

The sixteen-month window reshapes these matters more than anything else. If the loss began more than sixteen months before counsel was retained, Google's own record no longer exists, and no subpoena fixes that because the data is not being withheld. The case is then built from server logs, the client's analytics, archived captures, and third-party estimates whose limitations belong on the exhibit.

The evidence that establishes what changed on the site

Causation requires a mechanism, and it is almost always a change to the site with a date attached. The records that supply it:

• URL Inspection results from Search Console, reporting the fields an opinion turns on: lastCrawlTime, coverageState, googleCanonical and userCanonical (Google's chosen canonical URL versus the one the site declared), pageFetchState, robotsTxtState, and indexingState. The API returns these as enumerated values, which makes an exhibit built on them reproducible.
• robots.txt history — the file at the root of a site telling crawlers which paths they may request — with the history of <meta name="robots"> tags and X-Robots-Tag headers, which control whether a page may be indexed.
• Redirect maps and observed HTTP status codes, including whether a 301 permanent redirect was used where one was required, and whether chains or loops were introduced.
• Sitemap submission history, and CMS revision history and deployment logs with author and timestamp.

A single noindex directive left on a template after a staging deploy, dated to the day, is worth more to a causation opinion than any amount of narrative about declining rankings. It is a mechanism, it is documentary, and it ties to specific URLs whose traffic can be measured separately.

The evidence that establishes what changed at the engine

The defense's strongest material is often not about the site at all. It is about what Google did in the same window.

Google publishes a ranking-updates history with its own start dates and durations, and those windows are long. The March 2024 core update ran forty-five days and began on the same day as the March 2024 spam update. The December 2025 core update ran eighteen days. When a decline begins inside one of those windows, the claimant has to do more than point at the chart, and an expert who does not check the dashboard first will be shown it in deposition.

The other engine-side artifact is the Manual Actions report. A manual action is a human decision by a Google reviewer, recorded with a named category and a date — close to documentary proof that something violated a published policy. Its absence does not prove the opposite — but it does mean any claim of penalty is an inference rather than a finding.

Apply this honestly in both directions. A model that omits market events both sides agree were not caused by the defendant is the failure the Eighth Circuit described in Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8th Cir. 2000), where the model ignored inconvenient evidence. The analogue here is exact.

The evidence that establishes value — and where this page stops

Once a decline and a mechanism are established, the claim still has to show the lost traffic was worth something. The records are commercial rather than technical:

• Order and CRM data joined to landing page and to source or medium
• Historical conversion rates by page type, measured before the incident
• Paid search cost data as a substitution benchmark, where the same queries were purchasable

Notice what those have in common: none lives in Search Console, and none is an SEO artifact. This is where the search expert's work meets a financial expert's, and the boundary should be drawn explicitly in the report.

In my experience the cleanest division is this: the search expert establishes which URLs lost which volume of which measured quantity, over which dates, and why; the financial expert converts that into money. When one witness tries to do both, the weaker half contaminates the stronger. The method for the conversion — comparison period, benchmark, seasonality adjustment, treatment of concurrent updates — is a separate subject with its own reliability standards, treated on its own pages.

The ways these claims come apart

Every item below is one I would expect competent opposing counsel to raise.

1. The window has rolled off. Search Console holds sixteen months, so counsel retained eighteen months after the event has no engine-side record of the before-period.
2. The metric was never defined. Sessions, users, clicks, impressions, pageviews, and entrances are not the same quantity.
3. A tracking change is mistaken for a traffic change. An analytics migration, a consent deployment, a tag misfire, a bot-filtering change, or a time zone change produces a cliff that is an artifact of measurement.
4. The confound is a core update. The decline starts inside a documented update window.
5. Several things changed at once. A redesign, a migration, and a content pruning in the same month cannot be separated without URL-level data.
6. Position is used as a proxy for traffic. Rank trackers measure from data centers and locations that are not the client's users, and Search Console's average position is an impression-weighted average, not a rank.
7. An industry click-through curve replaces the site's own data. Published curves aggregate query types and verticals that may look nothing like this site.
8. Value is asserted rather than measured. Queries that historically converted at a fraction of a percent do not support a revenue claim at the site average.
9. Mitigation is absent. No reconsideration request where a manual action existed, no redirect repair, no disavow where a link attack was alleged.

Two are where opinions get excluded rather than merely weakened. Substituting the witness's confidence for an analytic method is the failure the Seventh Circuit described in Zenith Electronics Corp. v. WH-TV Broadcasting Corp., 395 F.3d 416 (7th Cir. 2005): reliable inferences depend on more than say-so. A method invented for the case is the failure in LifeWise Master Funding v. Telebank, 374 F.3d 917 (10th Cir. 2004), where the model was unique to this case and not based on any recognized standard.

The AI Overviews complication, and why it limits both sides

Any traffic loss claim filed now runs into a defense that part of the decline is attributable to changes in how Google presents results. It is a serious argument and deserves a precise answer.

Google states that AI Overviews appear only when its systems determine the format is additive to classic Search, and that they often do not trigger. More importantly for measurement, Google states that AI Overview and AI Mode traffic is included in the overall search traffic reported in Search Console, within the Web search type, and is not broken out separately.

Read that carefully, because it limits both sides symmetrically. A claimant cannot use Search Console to isolate the impact of AI Overviews and subtract it. A defendant cannot use Search Console to prove the loss was caused by AI Overviews rather than by its own conduct. The same fact constrains both arguments, and an expert who invokes it in one direction while ignoring the other is presenting an incomplete picture of the tool. Google's documentation on AI features is the source, and its version date should travel with the citation.

Frequently Asked Questions

Published 2026-08-11