Does meta description still matter? What CTR data shows

In February I rewrote 312 meta descriptions for a B2B client and average CTR went up 0.8 percentage points. In March I did the same on a fashion e-commerce and CTR dropped 0.3. Same methodology, opposite results. When I ran the consolidated analysis on 1,847 URLs in BigQuery pulling Search Console, it became clear that 'does meta description matter?' is the wrong question. The right one is: matters for which query type, at which position, and against which competition? The data comes from four different tenants, all with at least 90 days of baseline before the test.

First fact: Google rewrites your meta description in 62.78% of cases when the query is informational, based on our sample cross-referenced with Portent's 2024 study updated. For transactional queries, that drops to 31%. So if you are optimizing a PDP, your tag has a solid chance of showing up verbatim in the SERP. On a blog post explaining 'what is a canonical tag', Google will pull snippets from the body. That completely changes the ROI of rewriting. Before touching any description, run the query report in GSC and cross it with the affected URLs, exactly as I suggested in Search Console: 7 underused reports and what to extract from them.

Second fact, and the one that surprised the team most: average position matters more than the text. At positions 1-3, well-written descriptions add on average 1.2 pp of CTR over the auto-generated version. At positions 4-10, the gain drops to 0.4 pp and falls inside statistical noise. Below position 10, irrelevant. This matches the CTR benchmark by position: updated 2026 data we published last month and reinforces the logic of prioritizing pages already on the first page. Rewriting the description of a page at position 23 is wasted time you should be spending on Content decay: spotting the posts quietly losing traffic.

Third: what works, works in a boring way. The winning patterns in our tests were three. Concrete number up front ('Analysis of 1,847 URLs shows...'), a question mirroring the user query ('Is migrating to GA4 worth it?'), and a time promise ('In 4 minutes you walk away with a checklist'). Avoid vague adjectives like 'complete', 'definitive', 'best'. Google's model that decides whether to keep or rewrite seems to penalize boilerplate, and anyone working on titles has known this since the experiments described in Title tags that convert: 7 patterns tested on real SERPs. Descriptions follow the same specificity logic.

Fourth point, technical: length matters less than what used to be taught. The mean of descriptions that survived in the SERP was 142 characters, but the standard deviation was 38. There are 78-character descriptions outperforming 158-character ones. Google truncates in pixels, not characters, and mobile cuts earlier. Use Mangools' tool or Screaming Frog with the pixel width filter to audit at scale. Include this in your How to audit on-page SEO without falling into guesswork flow alongside title, H1, and canonical. And remember that if the page has Canonical tags: common mistakes bleeding your organic traffic pointing wrong, optimizing the description is bailing water with a leaky bucket.

Fifth: AI Overviews changed part of the game. On queries where SGE shows up, organic CTR for positions 1-3 dropped on average 18% in our panel since January, and description barely moves the needle because the user already consumed the answer. The defense is not a better description, it is content structured to be cited. I covered the how in Content for AI search: optimizing for SGE and Perplexity and Featured snippets: how to structure content for position zero. For queries without SGE, especially transactional and navigational ones with clear intent, a well-crafted description still delivers measurable gains.

Practical takeaway: stop rewriting meta descriptions in bulk. Filter in GSC your URLs at positions 1-10, with more than 100 impressions/month, on queries where SGE does not appear. That subset is where rewriting pays. For the rest, let Google do its job. Meta description in 2026 is no longer an SEO lever, it is a conditional CTR lever, and the difference between the two is worth your next results meeting.