Average position in Google Search Console is calculated as a weighted average of every impression — across every keyword, every device, every country, every query variation. Add 50 new pages that rank at position 25, and your site-wide average position will drop (numerically worse) even if every single one of your priority keywords improved their ranking. The metric is mathematically correct and strategically misleading at the same time.
If you've ever stared at GSC and asked "why did average position drop when traffic went up?" — this is why. The fix is to ignore the site-wide average and look at position per keyword.
How GSC Calculates Average Position
From Google's own documentation: average position is the average position of your top-ranking URL for each query, weighted by impressions. So a query that fires 10,000 times at position 25 has 10x the weight of a query that fires 1,000 times at position 3 when calculating the site-wide average.
Practical implication: if you publish a new blog post that picks up 5,000 monthly impressions at position 22, those impressions get added to your average position calculation and pull the average toward 22. Your old high-performing pages didn't move at all. The number on the dashboard moved anyway.
A Concrete Example
Suppose your site had 100,000 monthly impressions averaging position 8, across 50 keywords. You publish 10 new pages targeting long-tail keywords. They start ranking in positions 18-25 and contribute 50,000 new monthly impressions.
| State | Total Impressions | Weighted Avg Position | What Happened |
|---|---|---|---|
| Before | 100,000 | 8.0 | Stable 50-keyword footprint |
| After (new content) | 150,000 | ~12.7 | New pages dragged avg down |
| Conclusion | +50% traffic | Looks like decline | Metric is misleading |
Traffic grew 50%. Original keyword positions didn't change. Average position went from 8 to 12.7. A leadership team looking only at the position metric would conclude SEO is getting worse. The opposite is true.
The Other Way Average Position Distorts
Query Mix Shifts
When your content gets discovered for new long-tail variations of an existing query, average position falls because long-tail queries typically rank lower than head terms. More keywords ranking is a good thing; the average position metric reads it as bad.
Device Mix Shifts
Mobile rankings are typically lower than desktop for the same keyword because mobile SERPs have more features displacing organic results. If your mobile impression share grows (which has been the trend across most categories since mobile-first indexing), average position falls without anything actually getting worse.
Country Mix Shifts
If you suddenly start ranking in a new country — say, a piece of content gets picked up in India and accumulates 20,000 impressions at position 18 — the new country's impressions are weighted in proportionally. Average position falls because of expansion, not decline.
Branded vs Non-Branded Mix
Branded queries almost always rank #1. If branded search volume grows (because of a marketing campaign or product launch), the weighted average improves — making SEO look like it's working when actually only the brand is. The reverse is also true: if branded search drops after a campaign ends, the average position metric worsens even if non-branded SEO performance held flat.
What to Use Instead
Per-Keyword Position Tracking
Filter GSC by individual high-priority queries and watch their position over time. The average position for "google search console export" specifically is a real metric; the site-wide average position is an average of unrelated keywords.
The Position Range Filter Trick
GSC has a feature that most users don't know exists: you can filter Performance data by average position with operators. Click "+ New" → Search type / Date / Country / Device — but also: Position.
The Position filter accepts:
Position is greater than— useful for isolating struggling keywords (e.g., position > 10 finds page 2 and beyond)Position is smaller than— isolates top-ranking content (e.g., position < 4 for top 3 results)Position equals— exact position matchesPosition does not equal— exclusion
Similar operators exist for Clicks and Impressions filters. Impressions > 1000 + Position > 10 finds your striking-distance opportunities without any regex at all — the highest-impression page 2 keywords ready to optimize.
Combine Filters for Real Cohort Analysis
Stack the filters: Position > 10 AND Clicks < 5 AND Impressions > 500. That's a cleaner definition of "struggling high-impression keyword" than any keyword-level review of the raw data. Use this combination for striking-distance audits, content gap analysis, and prioritization meetings.
What This Means for Reporting
If you report site-wide average position to stakeholders, you're almost guaranteed to have an awkward conversation at some point when the metric moves in the wrong direction for a non-SEO reason. Two safer reporting approaches:
- Report per-keyword position for a defined target list. Pick 20-50 high-value keywords and track their average position. Stable methodology, predictable movement.
- Report clicks instead of position. Clicks are the metric that actually maps to value. Position is a leading indicator; clicks are the result.
The 2025-2026 Bug Complication
Average position calculation depends on impressions, which were over-counted by GSC for 50 weeks between May 2025 and April 2026. Google confirmed the bug and chose not to fix historical data. The inflated impression counts affected which queries dominated the weighted average — distorting site-wide average position differently from how it would have calculated with clean impression data.
For analysis crossing this period: don't compare average position across the boundary. Use clicks (unaffected by the bug) and individual keyword positions (which weren't averaged across query mix).
The Done-For-You Version
The functional shift: from a single site-wide metric that responds to every possible confound (new content, query mix, device mix, country mix, brand mix) to per-keyword time series that respond only to actual ranking changes on the keywords you care about. The number you check stops being a noisy aggregate and becomes a clean signal.
Site-wide average position isn't useless — it's just measuring something different from what most people assume it measures. Use it for what it is (a weighted aggregate across your entire keyword footprint) and don't use it for what it isn't (a measure of whether your SEO program is working).