Search Console Data as Evidence

What the data is, and why nothing replaces it

Google Search Console is the reporting service Google provides to the verified owner of a website. Its Performance report is the only first-party record of how that site appeared in Google Search: impressions (the number of times a link to the site was shown on a results page), clicks, click-through rate (clicks divided by impressions), and average position (the mean rank of the site's highest-placed link when it was shown). The report can be grouped by query, page, country, device, search appearance, and date.

Every other source in a search dispute estimates what Search Console records. Third-party rank trackers sample results pages from their own locations. Clickstream panels extrapolate from a subset of users. Analytics platforms see visits after they arrive, not the searches that produced them. When a damages opinion rests on traffic from Google, Search Console is the closest thing to a business record of that traffic that exists, and an opinion that ignores it when it was available invites the obvious question on cross-examination.

That is exactly why its limits matter. A litigator who knows Rule 34 cold will tend to treat a Search Console export as a complete ledger. It is not. It is complete in some places, sampled in others, and dated in a time zone that rarely matches anyone's server logs.

The sixteen-month window, and what it does to the timeline

Google's help documentation states it in one sentence: Search Console keeps data for the last 16 months, and analytics reports built on Search Console data carry the same sixteen-month maximum. The window rolls forward every day. Each morning, one more day of the record falls off the far end and cannot be retrieved from Google again.

The default view is shorter still. When a user opens the Performance report, it shows the past three months unless someone changes the date range. In my experience this is the most common production defect in search cases: a client is asked for its Search Console data, exports what is on the screen, and produces three months of a sixteen-month record. The export looks complete. It is the default.

The consequence is arithmetic. If a traffic loss began more than sixteen months before anyone preserved the data, Google's record of the period before the loss no longer exists unless someone exported it at the time or configured a scheduled export. That is not a discovery dispute a subpoena resolves. It is a gap in the universe of evidence, and the analysis has to be built from what remains: server logs, the site's own analytics, archived captures, and third-party estimates whose limitations have to be stated.

Google offers a scheduled daily bulk export of performance data to BigQuery, its cloud data warehouse, which keeps the data for as long as the account holder retains it. Configured before a dispute, it is the strongest preservation mechanism available. Google's documentation does not clearly say whether it fills in history from before the date it was configured, so treat it as prospective unless the account itself shows otherwise.

Rows, anonymized queries, and why the totals do not add up

Three properties of the report produce more disagreement between experts than anything else, and Google documents all three.

• The table is capped. The report table displays a maximum of 1,000 rows, so rare and long-tail rows can be omitted from the table while still being counted in the chart total. The Search Analytics API, the programmatic interface to the same data, returns up to 25,000 rows per request, with a default of 1,000.
• Some queries are withheld. Google omits rare queries from the query table to protect user privacy and calls them anonymized queries. They are left out of the table but included in the chart totals unless a query filter is applied.
• Only the important rows are kept. Google states that, because of internal limitations, Search Console stores and shows only the most important data rows. Filtered totals for matching and non-matching rows may not add up to the unfiltered total, and adding a page or search appearance filter changes how the data is aggregated, which can raise the reported totals.

The common mistake follows directly. An expert exports the query table, sums the clicks, and treats the sum as the site's search traffic. The sum falls short of the chart total by an amount that depends on how much of the site's traffic comes from rare queries, and that share can differ between the period before an event and the period after it. A before-and-after comparison built on summed rows can therefore show a decline that is partly a change in query mix. Compare totals with totals and rows with rows, and say in the report which one the opinion uses.

Dates, preliminary data, and matching one source to another

Search Console reports dates in Pacific Time. Server logs are usually written in UTC or in the server's local time, and an analytics property reports in whatever time zone it was configured with. A day in one source is not the same twenty-four hours as a day in another. For a monthly trend the difference is noise. For an opinion that ties a ranking change to a specific deployment on a specific day, it can move the event across a date boundary, and opposing counsel will find it.

The newest data is also provisional. Google labels recent figures as preliminary, still being collected and subject to change over the next few hours, and states that collected data is usually available in two to three days. The API exposes the same distinction through a data-state parameter that returns finalized data by default and fresh, incomplete data only on request. An export taken the day after an event and never refreshed may not match a later export of the same dates. Record the export date and the data state, and re-export before a report is served.

AI features counted inside ordinary search totals

For most of the period now in litigation, Search Console counted impressions and clicks from Google's generative AI features inside ordinary web search totals. The UK Competition and Markets Authority described the position plainly in a January 2026 consultation: publishers could not view separate metrics for their content in generative AI features, and clicks and impressions from those features were aggregated across Google Search as a whole within Search Console.

For any period before Google introduced separate generative AI reporting, a Search Console export cannot by itself separate AI Overview exposure from ordinary listings. An opinion that attributes a loss to AI Overviews using Search Console alone relies on a distinction the data did not record for that period. The separation has to come from dated captures of the results pages for the claimant's own queries, with the foundation questions those captures carry.

Who can produce the data, and what an agency dispute does to it

Access to a Search Console property is set by permission level. A verified owner has proved ownership with a token, such as an HTML file uploaded to the site or a DNS record. A delegated owner was granted ownership by a verified owner without a token and holds the same permissions. Full users can view all data and take some actions; restricted users have view rights on most data.

If every verified owner is removed, Google's documentation states that the remaining users lose access after a grace period, and that data continues to be collected but nobody can see it until someone verifies ownership again. Verification tokens stay on the site until someone deletes them, so a removed owner whose token is still in place can re-verify.

This matters most in agency termination disputes. The agency may hold the only verified ownership while the client's own HTML verification file or DNS record still sits on the client's server. Whether the client could restore its own access, and from when, is a fact to establish early. Whether that put the data within the client's control for preservation purposes is a question for counsel, but it turns on a technical fact an expert can establish.

Analytics retention is a setting, not a constant

Google Analytics 4, Google's current analytics platform, keeps detailed data according to a setting chosen by the property administrator. Google documents the options: for standard properties, user-level data can be retained for two months or fourteen months, and other event data for two or fourteen months; Analytics 360 adds twenty-six, thirty-eight, and fifty months for event data. Data that reaches the end of its retention period is deleted automatically in a monthly process. When a standard property becomes a Large property, event-level retention is reduced to two months automatically and older event-level data is permanently deleted.

Two further statements in the same documentation decide arguments. First, the retention setting does not affect standard aggregated reports. An opposing expert who says the analytics data is gone because retention was set to two months is usually overstating it: monthly sessions and channel totals in the standard reports survive, and what is lost is user-level and exploration-level detail. Second, a change to the retention period takes effect after twenty-four hours, during which it can be reversed, and a reduction deletes the affected data in the next monthly process. A reduction made after a preservation obligation attached therefore carries its own timestamps, which is the kind of fact a technical reconstruction can establish.

Do not assume what a property was set to, and do not assume a default. Check the setting, record it with the date you checked, and state in the report what level of detail was available for each period of the analysis.

Bing and Business Profile data are not substitutes

Bing Webmaster Tools, Microsoft's equivalent service, extended its search performance history from six months to sixteen months in October 2024, according to Microsoft's own announcement. Reports of a further extension have appeared only in trade press, so the sixteen-month figure is the one to rely on unless the account shows more.

Google Business Profile, the listing that appears in Google Maps and local results, reports impressions on a different basis. Google's own metric definitions count multiple impressions by a unique user within a single day as one impression. Search Console does not deduplicate that way, so the two impression counts cannot be added together or compared directly. Google also does not document how far back Business Profile performance data is available, which is a reason to export it on a standing schedule rather than assume it will be there when a dispute arrives.

Frequently Asked Questions

Published 2026-10-02