Google Search Console deletes all data older than 16 months. Not archives it, not compresses it — deletes it, permanently. There's no way to recover keyword performance data from 2021 once it's gone. The deletion happens silently, on a rolling basis, with no notification.
For most sites, this is a non-issue right up until the moment it isn't. The situations where it becomes genuinely consequential are: algorithm update investigations requiring historical baselines, year-over-year analysis on sites with strong seasonality, and any situation where you need to demonstrate multi-year organic traffic growth.
Why 16 Months
The 16-month limit appears intentional rather than arbitrary. SEO Stack's analysis suggests it was designed to enable full year-over-year comparison within the native tool — 12 months of current data plus a 4-month buffer for the comparison period. That makes sense as a design choice for the primary use case.
The scale of data involved is also relevant: the same analysis estimates a single site's GSC export can reach 656.9 million rows and 67 gigabytes. Google providing unlimited free historical storage for this volume of data would be economically unrealistic.
Third-party SEO tools like Ahrefs and Semrush are not viable substitutes for actual historical GSC data. Their keyword estimates carry approximately 48-50% error margins compared to actual GSC figures. They track ranking positions, not the actual clicks and impressions that GSC measures. They're useful for competitive intelligence but not for historical traffic reconstruction.
The BigQuery Solution
Google provides a native export from GSC to BigQuery — the same BigQuery dataset infrastructure described in Google Cloud Console. Once enabled, the export runs daily and accumulates data indefinitely. There's no export of historical data before the setup date; you can only preserve what's current and future.
This means the value of setting up the export is entirely dependent on when you start. If you set it up today, you'll have two-plus years of data by mid-2028. If you set it up in 2020, you have data going back to 2020. If you never set it up, you have 16 months and nothing before.
The correct time to set it up is immediately, because the data you lose every day is gone permanently.
What You Can Do With Historical Data Beyond 16 Months
True Year-over-Year Comparisons
The 16-month window lets you compare this month to the same month last year — but only barely. If you're looking at May data and it's already late May, the prior-year May data may be right at the edge of the retention window. For any month-over-month comparison involving periods further back than 12 months, the BigQuery export is the only option.
Algorithm Update Impact Analysis
Google ran four core updates in 2024 (March, August, November, December). If you want to understand the before-and-after impact of the March 2024 Core Update on your specific keywords — comparing six months before to six months after — 16 months of data gives you enough window. But for the March 2023 Core Update or anything before May 2024, the data doesn't exist in GSC anymore unless you have it in BigQuery.
Multi-Year Executive Reporting
Board reports and investor updates often require 3-5 year trend data. "Organic traffic grew 74% between 2021 and 2024" is a statement you can only make if you have data from 2021. Without BigQuery export, this kind of longitudinal reporting is impossible.
Seasonality Baseline Analysis
For businesses with strong seasonality — e-commerce, tourism, tax services, anything with a distinct peak period — understanding whether this year's peak is better or worse than previous peaks requires comparing same-season data across multiple years. Two or three years of BigQuery data makes this analysis routine.
Setting Up the Export
The setup process involves connecting your GSC property to a BigQuery dataset in Google Cloud Console. Full instructions are covered in our separate guide on How to Export Google Search Console Data to BigQuery. The short version:
- Create or select a Google Cloud project with BigQuery enabled
- In GSC → Settings → Bulk Data Export → connect to BigQuery
- Grant the required IAM permissions to allow GSC to write to your dataset
- Data begins flowing the next day and accumulates from that point forward
The Data You're Losing Right Now
If your site has been active since before May 2025, GSC has already deleted your pre-May-2024 data. Every day that passes without a BigQuery export running, the oldest surviving data ages out.
The specific loss: if you don't have an export running and an algorithm update hits in August 2026 that affects your site's traffic, you won't be able to compare your current performance to the pre-update baseline from more than 16 months earlier. You'll know traffic changed; you won't know the full historical context of where it changed from.
A Note on the 2025 GSC Impression Bug
For any historical data spanning May 2025 through April 2026: GSC over-counted impressions for 50 weeks during this period. Clicks were not affected. If your BigQuery data includes this period, impressions and CTR figures from that range need to be treated as approximate rather than reliable. Google confirmed the bug and declined to fix the historical data. For this period specifically, clicks are the reliable metric.
The practical implication for historical analysis: if you're building a dashboard that shows multi-year keyword performance, flag the May 2025 – April 2026 impression data as requiring interpretation. The clicks data from the same period is unaffected and usable.
The Compounding Value of Starting Early
Historical data has asymmetric value — it becomes more useful over time. The BigQuery export you set up today will enable analysis in 2027 and 2028 that you literally cannot do any other way. Setting it up two years from now restarts the clock; setting it up today means two more years of history when you need it.
The 16-month window in native GSC is adequate for most operational SEO work. It's the strategic, longitudinal analysis — the work that demonstrates compound program value — where the limitation binds. That's the analysis that tends to matter most when justifying budget, explaining traffic changes to leadership, or understanding the long-term impact of algorithm shifts on a site's organic presence.