Meta description

When Google decides your written description matches the query well, it uses it verbatim (subject to length constraints). When it decides another passage on the page answers the specific query better, it swaps in that passage instead — often bolding the matched query terms. Analysis of large SERP samples consistently finds Google rewrites the on-page description on a majority of results, particularly for long-tail and question-style queries.

This means your single written meta description has to work as a generalist pitch for the page's primary intent, while the page's actual body copy needs to independently contain clear, extractable answers to the secondary intents a title alone can't cover.

As AI Overviews and zero-click answers absorb informational queries, the traffic that still clicks through to a website is increasingly commercial or navigational — someone who has already decided they want to visit a site, not just get a fact. That remaining click is more valuable and more competitive, and the difference between a generic tag and a specific, benefit-led one is the difference between winning that click and losing it to the result below you.

Generative engines like AI Overviews and Perplexity also pull passage-level text, sometimes including meta description content, when constructing summarised answers with citations. A precise, well-written description increases the odds your summary sentence is the one that gets lifted.

Google truncates by pixel width, not character count, so wide characters (W, M, capital letters) eat the budget faster than narrow ones (i, l). As a practical rule of thumb, 150–160 characters is a safe ceiling on desktop; mobile SERPs sometimes show more, sometimes less, and truncation behaviour changes without notice.

Front-load the value proposition and the specific service or answer in the first 110 characters, because that is the portion most consistently rendered across surfaces.

The most common failure is writing the description once for the homepage and then leaving every other page on the default CMS fallback — often the first sentence of the body copy, unedited, which reads as robotic and unpersuasive.

Start from the query, not the page. List the top three or four real search queries the page should win, pulled from Search Console or keyword research, and write the description to speak directly to the searcher's intent behind those queries.

Include one concrete differentiator — same-day service, a specific price range, a guarantee, a service area, a credential — because generic authority claims ('trusted', 'leading') are ignored by searchers who have seen the same words on every competitor's listing.

End with a soft, specific call to action ('See pricing', 'Check availability in Austin') rather than a generic 'Learn more', and preview the rendered pixel width before publishing.

For an Austin-area service page — say, 'AC repair in Cedar Park' — the description should name the service area explicitly, reference response time or availability, and avoid wasting characters on brand boilerplate that belongs in the title tag instead. 'Same-day AC repair in Cedar Park & North Austin. Licensed techs, flat-rate pricing, 24/7 emergency calls.' does more work than a generic 'We are Austin's trusted HVAC experts.'

Because local intent queries are frequently long-tail and question-shaped ('how much does AC repair cost in Austin'), Google is especially likely to swap in a body-copy passage instead of the written tag — so make sure the page itself directly answers pricing and availability questions in the first two paragraphs, not just in the meta tag.

Search Console's Performance report gives you impressions, clicks, and CTR by page and query. A meta description rewrite should be evaluated against CTR at a stable average position — comparing CTR before and after a rewrite is only meaningful if ranking position didn't shift materially in the same window, since position itself heavily drives CTR independent of copy quality.

Run rewrites in batches by template or category rather than one page at a time, so you have enough impression volume to draw a statistically meaningful conclusion within four to six weeks.