How to write meta descriptions in 2026: AI Overviews changed the rules

Meta description optimization used to be simple: write 155 characters, include your target keyword, make it click-worthy. That advice still circulates on most SEO blogs in 2026, and it still gets repeated in audits we see from clients who’ve worked with other agencies. The problem is that Choco Media started noticing something in client accounts around mid-2025 — meta descriptions were being overwritten by Google far more frequently, and the reasons why weren’t random. They followed a pattern. If you’re still treating meta descriptions as a simple copywriting task, this post will change how you approach them.

This is for SEO practitioners, content leads, and marketing managers who own organic performance. Whether you’re managing a single brand site or a portfolio of clients, the principles here apply. We’ll cover what actually controls whether your meta description shows up, how Google AI Overviews shifted the calculus, and the format patterns that consistently outperform hand-written descriptions in 2026.

What you’ll leave with: a clear framework for writing meta descriptions that perform in classic SERPs, a specific format for improving your odds in AI Overviews, and an honest assessment of when writing meta descriptions is worth your time at all.

Why Google rewrites your meta description (more than ever)

Google has always reserved the right to rewrite snippets. But the rate at which it happens has climbed sharply. Research from Portent and Semrush in late 2024 found that Google rewrites meta descriptions in roughly 70% of cases for informational queries — up from around 62% in 2022. For navigational and transactional queries, the rate is lower, but still above 50%.

The rewrites happen for a predictable set of reasons:

• The meta description doesn’t match the searcher’s query intent closely enough
• The page body contains a more relevant answer than the description
• The description is too short, too long, or repeats the title verbatim
• The description is keyword-stuffed rather than informative
• Google determines the query warrants a specific extract from the page body

This doesn’t mean meta descriptions are pointless. When Google does use your written description, click-through rates on well-written ones are meaningfully higher. The goal is to write descriptions that (a) survive the rewrite filter where possible, and (b) work well on the pages where Google consistently honours them.

Where Google tends to honour your description

Brand and navigational queries almost always show the written description. Branded product and service pages do too, as long as the description is specific and non-generic. The pattern we see in client accounts: generic descriptions (“We’re a full-service marketing agency delivering results”) get rewritten almost every time. Specific, query-matched descriptions (“AI-first marketing agency in Rovaniemi — content, paid media, SEO”) survive far more often.

How AI Overviews changed the meta description equation

AI Overviews — Google’s generative summaries that appear above organic results — don’t use meta descriptions as their source. They pull from page content, structured data, and signals Google uses to assess authority on a topic. But their presence changes how meta descriptions function in SERPs in two ways that matter.

First, when an AI Overview appears, organic results are pushed down. The description text needs to work harder to earn the click for users who scroll past the AI summary. Generic descriptions fail here. Descriptions that signal something the AI Overview didn’t cover — a specific tool, a step-by-step, a local angle — perform better.

Second, the pages that get cited inside AI Overviews are often pages whose content is structured for direct answering. If your page is structured that way, your meta description should reflect that structure too — signalling to the searcher that the page contains the clear, direct answer they’re looking for. This is a cohesion signal: alignment between description, title, and on-page content helps both rankings and click-through.

The meta description is no longer just a click-through tool. It’s a coherence signal — one part of a chain that includes your title, your structured data, and your on-page content format. When they’re consistent, Google has less reason to overwrite any of them.

The format that performs in 2026

Based on what we observe in client accounts and what’s consistent with current Google guidance, here’s the format that holds up:

• Length: 140–155 characters. Not because it’s a hard limit, but because it’s consistently within the display window across devices and query types. Going shorter often gets rewritten; going longer gets truncated.
• Structure: Problem or outcome → brief method signal → trust or specificity marker. One sentence or two very short ones.
• Keyword placement: Target keyword in the first half of the description, not bolted on at the end.
• No generic openers: Avoid “Learn how to…”, “Discover the best…”, “Find out why…” — these are overused and signal nothing specific to Google or the user.
• Active voice: “We cover the 8 format patterns that survive Google’s rewrite filter” beats “Meta description patterns are explored in this guide.”

Worked examples

Here’s the same page described three ways, moving from weak to strong:

1. Weak: “Learn about meta description best practices and how to optimize your website for better search rankings.” (Generic, no specificity, likely to be rewritten.)
2. Mediocre: “Meta description optimization tips for 2026. Includes examples and a checklist for better click-through rates.” (Better, but vague on the value.)
3. Strong: “Meta descriptions survive Google’s rewrite in ~30% of cases. Here’s the format pattern that improves your odds and what AI Overviews changed.” (Specific, includes a data hook, signals the angle.)

The strong version is more likely to be used as written because it’s specific enough to be useful, and it aligns with what the page actually delivers.

When writing meta descriptions is and isn’t worth your time

There’s a real cost-benefit question here. If Google rewrites 70% of meta descriptions for informational queries, spending an hour on meta descriptions for a 50-page informational content hub is often a poor allocation of time. Here’s how we think about it:

• Always worth writing carefully: Brand pages, service pages, product pages, landing pages with paid traffic, cornerstone content pieces.
• Worth a 10-minute pass: High-traffic blog posts, category pages, pages that rank in the top 10 for competitive terms.
• Low ROI: Long-tail informational posts, pages not currently ranking in the top 20, posts where the body content is already highly structured (Google will extract a better snippet anyway).

The practical approach: use a template for low-priority pages, invest real copy time on high-value pages, and monitor which description styles survive rewrites in your Google Search Console click data.

Meta descriptions and AI answer engines (beyond Google)

ChatGPT, Perplexity, and other AI answer tools don’t read meta descriptions at all. They parse your page body, your structured data, and your heading structure. So if your primary goal is citation in AI answers, time spent on meta descriptions has zero direct effect. This is worth saying plainly because we see content teams conflate the two.

What does help both Google organic and AI citation is the same thing: clear structure, answer-first content, specific data points, and proper schema markup. Our post on SEO services covers how we approach this as part of a broader technical and content strategy. And if you want a deeper look at the GEO side of things, our guide on AI SEO in 2026 covers how answer engine ranking works across ChatGPT, Perplexity, and Google AI Overviews.

Schema markup and its relationship to snippet control

One underused tactic for improving snippet control: using structured data to signal the correct description to Google. The description property in Article, WebPage, and Product schema isn’t the same as the meta description tag, but when they match or align closely, you’re sending a consistent signal across multiple channels.

More importantly, FAQ schema (using FAQPage markup) creates a secondary snippet opportunity — the accordion-style result that shows Q&A pairs beneath the main result. These are distinct from the meta description but they give you more real estate in the SERP and more control over what appears. Pages with well-implemented FAQPage schema see higher CTR on average, and the structured format also helps with AI Overview citations.

• Use Article schema with a matching description property on blog posts
• Use WebPage schema on service and landing pages
• Add FAQPage to any content-heavy page that answers multiple questions
• Ensure your meta description, schema description, and OG description are all consistent

Monitoring which descriptions survive: a simple GSC workflow

Most teams write meta descriptions and never check whether Google is actually using them. Here’s a lightweight process for understanding what’s happening in your own account:

1. In Google Search Console, go to Search Results → click on a specific page
2. Open that page in a separate browser window and search for the exact query it ranks for
3. Compare the live SERP snippet with your written description — are they the same?
4. If Google rewrote it, look at what it used instead. Usually it’s extracting from the opening 100 words of your body content or from a specific answer passage
5. Decide: is the rewrite better or worse? If it’s better, leave it alone. If yours would be better, revise the description to be more specific and query-matched

This process takes about 15 minutes per high-priority page. Doing it quarterly for your top 20 pages will tell you more about snippet behaviour than any blanket rulebook.

For teams running structured content audits, this fits naturally into the kind of AI-SEO content audit process we’ve written about — checking meta descriptions is one of the 12 checks on that list, and this workflow gives you the practical method for executing it.

The honest summary on meta description optimization in 2026

Meta descriptions still matter — but their role has shifted. They’re less about driving clicks directly and more about coherence signalling: telling Google that your page is exactly what it says it is. The pages where your description survives are exactly the pages where coherence is highest. Getting that coherence right — consistent keyword placement, specific language, alignment with title and body — is the work.

AI Overviews haven’t killed meta descriptions. They’ve made the click harder to earn for some query types, which means the description needs to do more work when it does appear. The format advice above isn’t new, but the context has changed: generic descriptions used to be a missed opportunity. Now they’re a competitive disadvantage on any SERP where an AI Overview takes the top position.

• Write carefully for brand, service, and high-traffic pages; use templates for the long tail
• Match description, title, schema, and body — consistency is what survives Google’s rewrite filter
• Use FAQPage schema to earn a second snippet position on content-heavy pages
• Monitor what Google actually uses in Search Console quarterly; adjust based on what you find
• Don’t confuse meta description work with AI answer citation work — they require different tactics

If you want to work through this as part of a broader content or SEO strategy, we’re happy to look at your specific situation. Reach out here and we’ll schedule a conversation.