Your Search Console Data Expires in 16 Months

Search Console keeps 16 months of performance data. Past that, it is gone — not archived, not available on request, not recoverable by paying anyone. Deleted.

Sixteen months sounds like plenty. It is not, and the reason is a piece of arithmetic that catches people out.

Why 16 months is less than it sounds

The number is chosen so you can *just* compare a month to the same month last year — 12 months back, plus a few months of runway. That is the entire design intent, and it is a thin margin.

It means you can never answer a two-year question. Not 'is this Black Friday better than the one before last'. Not 'what did the seasonal curve look like before we redesigned'. Not 'has our traffic actually grown over two years, or did we just have a good quarter'. Those questions require 24 months, and Google will only ever hand you 16.

What you lose, specifically

• Multi-year seasonality. One year of data shows you a shape. Two shows you whether it is a *pattern* or a fluke — and only the second one is actionable.
• Pre-migration baselines. Redesigns and replatforms are judged against 'before'. If 'before' fell out of the window, you have no way to prove the migration helped or hurt, and the argument gets settled by whoever is most confident.
• Long-tail query history. The individual queries that quietly compound into most of your traffic. You cannot reconstruct them later; nobody else has them.
• Evidence. Any claim starting 'traffic is up since we…' is unfalsifiable once the baseline expires. This matters more than it sounds if you ever need to defend a decision.

The fix is boring and takes an afternoon

Export the data yourself, on a schedule, into somewhere you own. That is it. There is no clever solution and no product that can retrieve what Google has already deleted — the only version of this that works is the one you start *before* you need it.

The lazy version, today

Open Performance, set the range to the full 16 months, and hit Export. Do it for Queries and for Pages. Drop the CSVs somewhere durable. This takes four minutes and is strictly better than nothing — which is what you have right now.

The version that actually works

The Search Console API returns far more than 1,000 rows, and can be run on a schedule. The shape of it:

POST https://www.googleapis.com/webmasters/v3/sites/{siteUrl}/searchAnalytics/query

{
  "startDate": "2026-06-01",
  "endDate": "2026-06-30",
  "dimensions": ["date", "query", "page", "country", "device"],
  "rowLimit": 25000,
  "startRow": 0
}

Page through with startRow until you get fewer rows back than you asked for. Write the results to BigQuery, Postgres, or honestly just dated Parquet files in object storage — the storage choice matters far less than the fact that the job runs at all.

Run it monthly. Set it to fetch the month that ended at least three days ago, so the data has settled. That is the whole system.

Or let Google do it: the BigQuery bulk export

Search Console has a bulk data export that streams your data into BigQuery daily, with no row limits and no anonymization beyond Google's usual privacy filtering. You configure it once in Settings and it accumulates from that day forward.

Two caveats worth knowing before you assume this solves it:

• It is not retroactive. It starts collecting the day you turn it on. It cannot recover the 16 months you already have, let alone what has already expired.
• BigQuery costs money. Not much at this volume, but it is a billed GCP project, not a free Search Console feature.

If you are going to do this, the best day to have turned it on was the day you launched. The second best is today — and that is not a motivational cliché here, it is just arithmetic. Every day you wait, one more day falls off the back of the window and is gone for good.

The point

Nobody notices this limit until the day they need the data, and on that day it is unfixable. It is one of the very few SEO problems where the entire cost is determined by a decision you make *before* the problem exists.

Go export your 16 months. It takes four minutes, and it is the highest ratio of future value to present effort available to you in the whole tool.