Write the description as a definition, not a pitch. Length, voice, and a before/after table.
Open a handful of AI answers that cite product sites and you will notice a pattern: the reused sentence often sounds like a definition, not a tagline. "Northstar Analytics is a B2B product analytics platform for mid-market SaaS" travels. "The #1 revolutionary way to 10x your insights!!!" does not.
Your meta description is not a guaranteed citation. It is one of the first short strings machines and humans skim when they need a one-line explanation. Write it like a definition you would put on a slide for a smart friend.
What a meta description is for in an AI context
In classic SEO, the meta description influences the Search snippet more than rankings. In answer-engine contexts, short descriptive text — from meta, from the opening paragraph, from schema description — is candidate phrasing for summaries and entity cards.
You cannot force reuse. You can stop giving models only slogan mush to work with.
| Job | Good meta description | Bad meta description |
|---|---|---|
| Define the entity | Who you are + category + for whom | Superlatives with no noun |
| Disambiguate | Distinct from a similarly named app | Shared nickname only ("NS") |
| Stay accurate | Matches the page today | Promises a feature you sunset |
| Stay portable | Reads fine without your logo | Depends on a campaign joke |
Length, voice, and structure
Aim for roughly one or two plain sentences. Search UIs still truncate; answer reuse prefers clarity over stuffing.
Voice rules that travel:
- Lead with the brand or product name when the page is about that entity
- Use the category word your buyers use ("product analytics", not "insight OS")
- Prefer present tense facts over future hype
- Skip "delve", "leverage", "game-changing", and "in today's digital landscape"
A simple pattern:
{Brand} is a {category} for {audience}. {One concrete capability or differentiator}.
Hypothetical example for Northstar Analytics:
Northstar Analytics is a B2B product analytics platform for mid-market SaaS teams. It connects product usage data to revenue dashboards without a dedicated data team.
Before / after table
| Page | Before (pitch) | After (definition) |
|---|---|---|
| Homepage | Unlock next-level growth with AI-powered magic. | Northstar Analytics helps SaaS companies see which features drive retention. |
| Pricing | Flexible plans for every stage of your journey!!! | Northstar Analytics plans start with a free scan-style trial tier for small teams; Agency includes seats for client work. (Adjust to your real plans — never invent prices in meta.) |
| About | We are passionate innovators disrupting industries. | Northstar Analytics is a remote-first software company building product analytics for B2B SaaS. |
| Docs hub | Everything you need to succeed. | Documentation for installing Northstar Analytics, connecting data sources, and building dashboards. |
| Blog post | You won't believe these 7 tips. | How Northstar Analytics customers map activation events — definitions, examples, and pitfalls. |
Notice the after column still sells — it just sells with nouns.
How to rewrite your key templates
- List the 10 URLs most likely to be fetched (Home, About, Pricing, product, top docs).
- Paste each current meta description into a spreadsheet column.
- Highlight anything that lacks a category noun or the brand name.
- Rewrite using the definition pattern; keep claims you can defend on-page.
- Align
og:descriptionand, where you use it, schemadescriptionso you are not teaching three different stories. - Deploy and view source to confirm the live tag matches the CMS field (caching lies).
Relationship to titles and H1s
Meta description, title tag, and H1 should agree on the entity string. Clever titles that drop the brand already hurt entity clarity; a meta that invents a second nickname makes it worse. Run the three-field check in H1 brand consistency.
| Field | Must include | Soft include |
|---|---|---|
| Title | Brand or product name | Category |
| H1 | Same name family | Benefit line |
| Meta | Name + category | Audience + one fact |
Mistakes
Keyword stuffing.
"Best AI analytics AI tool AI dashboard AI SaaS" reads like spam to people and models.
Duplicate metas across 40 pages.
Every article saying "Welcome to our blog" wastes the slot.
Contradicting the first paragraph.
If meta says you serve enterprises and the H1 says freelancers, you taught a conflict.
Call-to-action only.
"Book a demo today" is not a description of what exists.
Unlabeled hypotheticals presented as facts.
If you must illustrate a metric, frame it as an example — same rule we follow in this blog's style for numbers.
How to verify
- Crawl the key URLs and export title + meta (Screaming Frog, site crawl, or a quick script).
- Spot-check that brand + category appear.
- Ask a colleague who does not work on the page: "What does this company do?" after reading only the meta. If they shrug, rewrite.
- Optional: compare whether assistants paraphrase your definitional sentence when they do mention you — directionally useful, not a KPI to overfit.
BrandKnown's readiness score checks website identity and access issues; it is not a citation forecast. A clear meta description is still cheap insurance for every system that skims your <head>.
Write the description as a definition, not a pitch. Length is secondary to whether a stranger could reuse the sentence without embarrassment.