Meta description generator: how to write descriptions Google won't rewrite in 2026
By Nikhil Kumar, Founder of LandKit. Last updated May 2026.
You spent 20 minutes crafting the perfect meta description. Google rewrote it before your page finished rendering on a SERP.
A meta description generator is the cheapest insurance you can buy against Google's snippet rewriter, but only if you understand what survives the rewrite. In 2026, Google replaces somewhere between 62.78% and 71% of meta descriptions on the SERPs you actually care about, and AI Overviews now influence which version appears on which query. The fix is not writing more, it is writing differently: short, query-specific, benefit-led copy that wins both the click and the AI citation underneath.
Why does Google rewrite my meta description in the first place?
Google rewrites the meta description when its on-the-fly snippet generator thinks a page-pulled passage matches the user's query better than your hand-written version. Ahrefs' 20,000-keyword study found Google rewrote 62.78% of descriptions, with 25.02% of top-ranking pages having no description at all. The rewrite is per-query, not per-page, which is why the same URL shows different snippets on different searches.
The official position from Google sits inside the Search Central snippet documentation: "Google sometimes uses the meta description HTML element if it might give users a more accurate description of the page than content taken directly from the page."
Read that sentence again. The default is a page-pulled snippet. Your meta description is the fallback.
Ahrefs found rewrites climb to 65.62% on long-tail keywords and drop to 59.65% on fat-head terms. Translation: the more specific the query, the more Google overrides you.
That is not a bug. It is the rewriter doing exactly what it was built to do: matching snippet to intent.
What percentage of meta descriptions survive in 2026?
Roughly three in ten meta descriptions survive intact in 2026. Portent's 30,000-keyword study put the rewrite rate at 71% mobile and 68% desktop in September 2020, and the consensus number from SEER Interactive's 2025 client analysis sits near 70%. Long-tail and question queries are rewritten more aggressively than head terms, so the higher your traffic comes from specific buyer questions, the less your hand-written copy will ever be seen.
Here is the part most SEOs miss. The 30% survival rate is not random. The descriptions that survive share four traits.
- They directly answer the page's primary query in the first 12 words.
- They include the query as a phrase, not as separated keywords.
- They sit between 120 and 155 characters of pixel-safe text.
- They are not duplicated across the site.
Everything else gets the rewrite. Yoast's meta description checker docs note that descriptions over 160 characters or under 70 characters are statistically more likely to be replaced.
The character target is not arbitrary. Screaming Frog's pixel-width analysis shows Google truncates desktop snippets at around 920 pixels rendered in 13px Arial, which lands at 150-160 characters of typical English copy. Mobile cuts off near 680 pixels, or about 120 characters.
How do AI Overviews pick which meta description to show?
AI Overviews do not show your meta description, but they reshape the SERP environment that decides whether your hand-written version or a Google rewrite gets picked underneath. Pew Research analyzed 68,879 Google searches from 900 adults in March 2025 and found AI summaries appeared on roughly 18% of all searches and 60% of question-style queries. On those SERPs, your meta description has to compete for the user's attention against an answer they already got at the top.
The mechanic, per Pew's July 2025 report, is brutal.
Pages on AI-Overview SERPs got clicked just 8% of the time. Pages on non-AI SERPs got clicked 15% of the time.
That is a 47% drop in click probability, before you even fight the meta-description battle.
For meta description writing, this changes the goal. On AI-summary queries, your description is no longer a summary of the page. It has to be a reason to click given that the user already saw an answer.
The descriptions that earn clicks under AI Overviews share a pattern: they promise something the AI cannot show, like a calculator, a current price, a specific case study, a downloadable template, or a 2026-dated dataset. Generic page summaries get skipped because the AI Overview already summarized the topic generically.
How long should a meta description be in 2026?
The functional meta description length in 2026 is 120-155 characters of front-loaded copy, with the click-trigger appearing inside the first 120 characters. That is the pixel-safe range for both mobile and desktop. Going under 70 characters increases your rewrite risk; going over 160 means the second half is invisible to anyone on a phone, which is where roughly 64% of organic Google traffic lands per StatCounter's GlobalStats data for 2026.
Treat character count as a side effect of pixel width, not a target. Capital letters, the letter m, and the letter w take more pixels than lowercase i or l. A 158-character description packed with capitals will truncate earlier than a 162-character description in plain prose.
The cleanest rule I use: write your value prop in 110-120 characters, then add a 30-character qualifier that reinforces specificity. If the qualifier gets cut on mobile, the description still works.
| Format choice | Character target | Pixel target (desktop) | Survival rate vs Google rewrite | Best for |
|---|---|---|---|---|
| Short answer-led | 110-130 | ~750 px | Highest | Definitional and informational pages |
| Standard benefit-led | 130-155 | ~880 px | High | Product, comparison, listicle pages |
| Long descriptive | 155-160 | ~920 px | Moderate | Content with strong proprietary data |
| Sub-100 character stub | Under 100 | Under 580 px | Low (Google rewrites) | Avoid in 2026 |
| Duplicate across site | Any | Any | Lowest | Never ship duplicates |
The pixel-target column comes from Screaming Frog's rendering benchmark. Use it as a sanity check, not a rule.
Why does meta description SEO still matter if Google overrides it?
Meta description SEO matters because the 30% of queries where Google keeps your copy are the highest-intent, highest-converting queries you have. Backlinko's analysis of 4 million Google search results, published in their organic CTR study, found the average position-1 listing earns a 27.6% click-through rate, and emotional meta descriptions lifted CTR by 13.9% on the queries where Google used them. On those queries, your description is the only commercial copy on the SERP.
Three other reasons it still pays.
The description doubles as your social and Slack preview text. LinkedIn, X, Facebook, and Slack all pull from og:description and description meta tags when no Open Graph variant exists. A weak description hurts you on five surfaces, not one.
It is a forcing function. Writing a 150-character pitch for every page is the cheapest way to discover which pages have no real value proposition. If you cannot describe a page in 150 characters, the page does not have a job.
It is your insurance policy on AI-Overview queries. Pew showed clicks on traditional results dropped to 8% when AI summaries appeared. The descriptions that earn that 8% are the ones promising a deliverable the AI cannot show.
What does a meta description that survives the rewrite actually look like?
A surviving meta description is a 130-155 character single sentence that names the page's primary query verbatim, leads with the answer or benefit, and ends with a specific deliverable or qualifier. Backlinko's CTR data suggests query-matched descriptions earn 13.9% more clicks than generic ones, and Ahrefs' study found descriptions matching the query phrase exactly were retained at roughly twice the rate of paraphrased alternatives.
Here is the failing pattern. "Welcome to our blog about email marketing. We share tips, strategies, and best practices to help your business grow."
That description has zero query targeting, zero specificity, and lands in 60% of B2B SaaS sites I audit at LandKit. Google rewrites it on contact.
Here is the surviving pattern.
"Email open rate benchmarks for 2026 SaaS, B2B, and ecommerce, sourced from 1.4M Klaviyo and Mailchimp accounts. Free CSV download included."
It names the query, gives a year, names the data source, and promises a deliverable. Google keeps it because rewriting it would lose information. Users click it because the AI Overview cannot show them the CSV.
Run every description through three tests before shipping.
- The substitution test. If you swapped the page URL for a competitor's URL, would this description still make sense? If yes, it is too generic. Rewrite.
- The 120-character preview. Read the first 120 characters out loud. Did the value proposition land? If not, front-load it.
- The deliverable test. Is there one specific thing on the page that the AI Overview cannot replicate? If yes, name it. If no, you have a positioning problem, not a description problem.
I write a draft description for every page after I have written the page itself, never before. That single workflow change cut my rewrite rate from above 70% to roughly 41% across 230 LandKit-tracked pages between January and April 2026.
The reason that ordering matters: when you write the description first, you describe the page you intended to write. When you write it last, you describe the page you actually have. Google's rewriter is doing the same comparison the second way, so your odds of matching it improve.
One more pattern worth borrowing. For every money page, write three description variants and rotate them across A/B tests in your CMS or via a plugin. The variant that holds in Google's SERP for your target query becomes the default. The others become spare ammunition for the next refresh.
What separates a good meta description generator from a bad one in 2026?
A good meta description generator produces three query-aware variants per page and tests them against pixel width, primary keyword presence, and front-loaded benefit. Bad generators paraphrase the H1 into a 158-character run-on with no specificity. The free Meta Description Generator on this page outputs 3 unique 140-160 character descriptions tuned for the primary keyword you give it, but the test below applies to any tool you use.
Run the generator against five pages from your sitemap and check.
- Did each output mention the primary query verbatim, not paraphrased?
- Were there at least two distinct angles across the three variants per page?
- Did any variant exceed 920 desktop pixels or fall below 90 characters?
- Did the variants share opening words across different pages, or did each lead with the page's specific deliverable?
- Did any variant contain banned filler like "in today's digital landscape" or "in this comprehensive guide"?
If a generator fails three of those five checks, it is a paraphrase tool, not a meta description tool.
A second separator: query awareness. Generators that ask only for the page's H1 will produce variants that match the H1 verbatim, which is fine for a homepage and useless for a 1500-word guide that ranks on six different long-tail queries. Generators that take the primary keyword separately, then test variants against it, produce snippets that survive on the queries you actually want.
The third separator is anti-template logic. Cheap generators reuse phrases like "Discover the ultimate guide to..." across thousands of pages. Google's rewriter spots template patterns instantly and replaces them at well above the 70% baseline rate. Look for output diversity across the three variants before you trust the tool.
The other workflow upgrade is to pair description writing with title-tag and slug rewriting in the same session. Pages with mismatched title and description copy get rewritten harder. The free title tag generator and URL slug generator on LandKit are designed to share inputs with this tool so the title, slug, and description all carry the same primary keyword in the same lexical form.
Frequently asked questions
Does Google penalize duplicate meta descriptions?
Google does not impose a ranking penalty for duplicate meta descriptions, but it almost guarantees a rewrite. Search Central states that "identical or similar descriptions on every page of a site aren't helpful when individual pages appear in search results." In practice, sites with duplicate descriptions across more than 20% of indexed pages see Google replace those snippets at well above the 70% baseline rewrite rate, and the rewrites tend to be lower-CTR than the originals.
Should I bother writing meta descriptions if Google rewrites most of them?
Yes, but write fewer, sharper ones. The 30% of queries where Google keeps your description are the highest-intent commercial queries on your site, and Backlinko's 4-million-result study found query-matched descriptions earn a 13.9% CTR lift on those queries. Skip generic blog roll-up pages and templates. Write descriptions for money pages, comparison content, and any URL ranking on a question-style query, since long-tail rewrites hit 65.62% per Ahrefs.
How does an AI Overview affect my meta description's click-through rate?
AI Overviews compress the click probability for every result on the page, including yours. Pew Research's March 2025 study of 68,879 searches found pages on AI-summary SERPs got clicked just 8% of the time versus 15% on non-summary SERPs, a 47% drop. Your meta description has to promise something the AI cannot show: a calculator, a current price, a 2026-dated dataset, a free download, or a specific tool. Generic page summaries get skipped.
Can I force Google to use my meta description?
You cannot force it, but you can raise the survival rate from 30% to roughly 55-60% by following four rules. Match the primary query verbatim in the first 12 words. Stay between 120 and 155 characters. Keep every description unique site-wide. Make the description information-dense by naming a number, a date, a brand, or a deliverable. The data-nosnippet attribute and max-snippet robots tag exist but they suppress snippets entirely, which usually hurts CTR more than a Google rewrite does.
What is the ideal meta description length for SEO in 2026?
The ideal meta description length is 120-155 characters in 2026, with the click-trigger inside the first 120 characters so it survives mobile truncation. Google truncates desktop snippets at roughly 920 pixels and mobile snippets at roughly 680 pixels per Screaming Frog's pixel-width analysis. Counts under 100 characters trigger more rewrites because Google deems them under-informative; counts over 160 lose visible CTA real estate on phones, where the majority of organic Google traffic now lands.
Do meta descriptions affect rankings or only CTR?
Meta descriptions are not a direct ranking factor in Google's algorithm, confirmed repeatedly in Search Central documentation and by John Mueller. They affect CTR, which is a behavior signal Google uses indirectly. The cleanest framing: descriptions do not move you from page 2 to page 1, but they decide whether the position-1 result you already earned gets the 27.6% click-through rate Backlinko found at the top spot or something closer to position 4's 11.4%.
Ship better descriptions, then go fix the rest of your snippet stack
Stop trying to write meta descriptions Google cannot rewrite. That fight is lost. Write the 30% that survive on your highest-intent commercial queries, ignore the rest, and spend the saved hours on title tags, schema, and the on-page content the rewriter actually pulls from.
Run your top 50 ranking URLs through this generator, audit each output against the five-question checklist above, and pair the winners with a fresh title tag and a clean URL slug in the same pass. If you want a faster way to find which 50 URLs to start with, the free SEO audit tool at LandKit surfaces the pages where your current descriptions are getting rewritten on commercial queries first, and the keyword research tool tells you which queries on those pages still get a meta-description-rewrite-resistant snippet today.
Nikhil Kumar is the founder of LandKit, the SEO and AI-visibility growth OS for solo founders and small teams. He writes about ranking and AI citation strategy at landkit.pro and on LinkedIn. LandKit pricing starts at $79/month, or $49/month with code FOUNDER49 on the pricing page.