Four Postgres queries caused 90% of our slow-response alerts. See how to diagnose each with EXPLAIN ANALYZE — with fixes that took under an hour.
Leads with the number readers will remember; pairs the method (EXPLAIN ANALYZE) with the time promise.
Free with an account · 5 angles, all under 160 chars
Get 3-5 distinct options under the hard 160-character SERP limit — benefit-led, question-open, stat-forward, how-to, brand-first — with character counts and a recommended pick.
See how it works — click any example
Benefit-led, question-open, stat-forward, how-to-spec, and brand-first. Every variant stays under the 160-character SERP limit (80 chars for CJK languages).
3-5 descriptions with verified char counts, a rationale per variant, a recommended pick, and 0-3 page-specific diagnostics. Pair with the Title Generator and FAQ Generator for a full on-page SEO pass.
Pages with no content — describe what the page actually does. Stuffed keyword lists — we write for humans + SERP. Brand names not provided by you — we never fabricate a brand.
Free with an account — 30/month. Upgrade to Pro for 1,500/month. Each run costs 1 credit.
Input: "Detailed guide for backend engineers on fixing slow Postgres queries. 4 causes, EXPLAIN ANALYZE examples, concrete fixes." · Page type: howto · Tone: direct · Keyword: fix slow postgres queries
Four Postgres queries caused 90% of our slow-response alerts. See how to diagnose each with EXPLAIN ANALYZE — with fixes that took under an hour.
Leads with the number readers will remember; pairs the method (EXPLAIN ANALYZE) with the time promise.
Fix slow Postgres queries in an hour — no new indexes, no re-architecture. EXPLAIN ANALYZE walk-through for backend engineers.
Reader gets the outcome first, plus the specific objections ("no new indexes") addressed inline.
Why is your Postgres query slow? 4 common causes, every one diagnosable with EXPLAIN ANALYZE. Fixes and code samples included.
Mirrors the exact question a backend engineer would search; immediate scope preview.
Step-by-step: diagnose slow Postgres queries with EXPLAIN ANALYZE. 4 common causes, concrete fixes, under an hour.
How-to format with verifiable specifics (4 causes, 1 hour) — pattern Google snippets favor for tutorial intent.
+ 3-5 target keywords to cross-verify with your H1 and URL · 0-3 page-specific diagnostics
Google's search snippet on desktop truncates at around 155-160 characters (about 920 pixels), and on mobile it's closer to 120. Anything past the limit gets cut off with an ellipsis. Write to 140-155 for a safety buffer on desktop; the shorter mobile limit is why the first 120 chars should carry the core pitch.
Yes — just not for ranking. Since 2009 Google has been explicit that meta descriptions don't affect search ranking, but they heavily influence CTR, which does. In 2024 Moz measured Google rewriting descriptions ~65-70% of the time, but a strong description (a) wins the render when Google does keep it, and (b) influences what Google extracts from the page when it rewrites. Either way: write a good one.
Every run returns options that pick different structural angles: benefit-led (what the reader gets), question-open (leads with the question the reader typed), stat-forward (leads with a number: "4 fixes that shaved 90% off query time"), how-to-spec (concrete outcome + time), brand-first (your brand as anchor, then benefit + CTA). You get real choices instead of surface rewordings of the same idea.
No — but you should if you have one. If you leave it blank, the model derives the strongest candidate keyword from your topic and uses it in every option. If you supply one, we fit it into every option naturally. The keyword shows up in the targetKeywords list so you can cross-check your H1 and URL carry it too.
Blog, product, landing, category, service, about, how-to, and listicle — each with a different formula. Product pages lean brand + benefit + CTA; landing pages get a CTA verb in every option; listicle pages want count + differentiator ("10 best coffee roasters in Portland — with roasting dates, not "fresh"-fresh claims."). Pick the page type that matches; the output adapts.
Only if you supply it. If you do, the "brand-first" option uses your brand as the anchor, and the brand may appear (usually at the end, separated by "·" or "|") in other options too. If you don't supply a brand, none of the options fabricate one.
The model picks one option as its recommendation based on the page type + tone you chose — usually the most balanced of the five. Use it as your default if you're deciding quickly. If you're A/B testing (worth doing on high-traffic pages), pick 2 options with different angles, ship both to different URL subsets or across a redeploy, and compare CTR in GSC.
Yes. Describe your page in any language and the model produces meta descriptions in that language. Japanese, Chinese, Spanish, French, German, Korean, Portuguese, Italian, Russian, and more. Character limits differ per language — CJK languages hit the pixel cutoff at fewer characters since each character is wider. The model stays under 80 CJK characters as a safe target.
Specific checks for the page itself — things that make the meta description actually land well. E.g. "Your title tag is 71 chars — will truncate on mobile, shorten to under 60" or "H1 doesn't contain your primary keyword — rewrite for topical alignment". Up to 3 per run, specific to your page, never generic SEO advice.
First 3 runs per day are free without signup. Free account gets 30 credits/month; each meta description run costs 1 credit. Pro plans have 1,500/month — enough for ~1,500 pages a month across a site.