> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getcited.in/llms.txt
> Use this file to discover all available pages before exploring further.

# Meta Description Quality

> How Cited measures whether your meta descriptions tell AI models — and the systems that summarize content for them — what each page is actually about.

AI models use meta descriptions to understand what a page is about before fetching its full content. Missing descriptions force AI to guess from prose; generic descriptions force the same guess. Quality descriptions get pages surfaced in answers; missing or off-spec descriptions don't.

## Methodology

Cited samples up to 5 pages from your site and checks the meta description on each against two criteria: presence and length.

**Presence.** Every sampled page either has a `<meta name="description" content="…">` tag with non-empty content, or it doesn't. Pages without descriptions force AI models to extract a summary from page text — slower, less reliable, and often wrong for landing pages or thin content where the first paragraph isn't summary-shaped.

**Length quality.** A description between 80 and 160 characters scores as "quality." This range matches the calibration used by AI summarizers and search engine snippet generators: too short and the description doesn't carry enough information for AI to differentiate the page from siblings; too long and the description gets truncated, often mid-sentence, losing the most informative trailing content.

The score is based on the **quality rate** across sampled pages — the percentage of pages with descriptions in the 80-160 character band:

* **80% or more pages have quality descriptions** → 6/6. Full score.
* **50-79% quality** → 3/6. Partial — most pages are right, but enough are off to drag the score down.
* **Any pages have descriptions, but few hit the quality band** → 1/6. Descriptions exist but aren't optimized.
* **No pages have descriptions** → 0/6. Critical failure.

The signal scores out of 6. Status thresholds: pass at 5/6, partial at 2/6, fail below.

Why average quality across pages instead of per-page check? AI citation favors consistency — a site where most pages have good descriptions signals editorial discipline, while a site with one perfect description and four missing signals neglect.

## Verification

You can verify our finding yourself in a browser.

**Step 1: Open the pages we sampled.** Cited reports the URLs we tested. Open each in a new tab.

**Step 2: Check the meta description tag.** Right-click the page, View Page Source (`Cmd+U` / `Ctrl+U`), and search (`Cmd+F` / `Ctrl+F`) for `name="description"`. The full tag looks like `<meta name="description" content="…">`. The content attribute value is what the scanner extracts.

**Step 3: Count the description length.** Copy the content attribute value (without the surrounding quotes) into any character counter — for example, paste it into the DevTools Console with `('…').length` substituting your description. The 80-160 character band scores as quality.

**Step 4: Spot-check the description content.** Length is necessary but not sufficient. A 100-character description that just repeats the page title doesn't help AI understand the page. Read each description and ask: would this description tell an AI model what's unique about this page versus a generic competitor page? If not, the description is technically passing this signal but not pulling its weight in AI citation.

If your verification disagrees with Cited's finding, that's a bug — let us know.

## Technical detail

The meta description tag is governed by the HTML Living Standard and has been a stable convention since HTML 4 (1997). Search engines have always used it; AI models inherited the same parsing path because the data is already in their training corpora and refresh cycles.

**Extraction logic.** Cited's scanner uses regex against the raw HTML (not DOM parsing) for this signal:

* Primary pattern: `<meta name="description" content="…">` with the `name` attribute before `content`
* Reversed pattern: `<meta content="…" name="description">` with `content` before `name`
* Both patterns are case-insensitive and tolerate arbitrary additional attributes (e.g., `id`, `lang`)
* The content attribute value is extracted between matched quotes (single or double)

The regex approach trades off completeness for speed — it handles the overwhelming majority of real-world meta tag formats but misses edge cases like multi-line content attributes or attributes with escaped quotes inside the value.

**Quality threshold.** Length is computed on the trimmed content string. The 80-160 character range comes from empirical AI summarization studies showing descriptions in this band are most likely to be quoted in AI answers verbatim. Above 160, AI models typically truncate; below 80, they extract additional content from the page body to fill in.

**Score calculation.** Quality rate = pages with quality descriptions ÷ total sampled pages. Present rate = pages with any description ÷ total sampled pages. The scoring tier is taken from the higher-value rate that the page hits.

**Edge cases the scanner handles:**

* **Multiple meta description tags on one page.** The regex finds the first match in document order. If a page has two `<meta name="description">` tags (which is invalid HTML but happens — usually one in `<head>` and one duplicated by a plugin), the first one wins.
* **Whitespace-only content attributes** — a description like `content="   "` is detected as present but the trimmed length is 0, so it fails the quality check.
* **Open Graph descriptions** — `<meta property="og:description">` tags don't satisfy this signal even though they serve a similar purpose. OG descriptions are scored separately under the Open Graph Tags signal.
* **Server-rendered vs client-rendered descriptions** — some single-page applications inject the meta description via JavaScript after page load. The scanner reads the post-hydration HTML (after Puppeteer's 3-second delay), so JS-injected descriptions are detected.
* **Encoded HTML entities** — descriptions containing entities like `&amp;` or `&#39;` count by character length including the entity, not the decoded form. A description with three `&amp;` substitutions reads as 18 entity characters where a user would see 3 ampersands.

**What this signal does not measure:**

* **Description content quality.** The scanner counts characters, not meaning. A page-title-repeat description scores the same as a unique value-prop description.
* **Description uniqueness across pages.** Every page can have the exact same description and score 6/6 if all are in the 80-160 band. Duplicate descriptions are a real signal of editorial neglect but aren't modeled here.
* **Keyword presence.** No keyword matching against page topics or brand terms.
* **Brand name inclusion.** Some style guides require the brand name in every description (e.g., "…— BrandName" suffix). The signal doesn't check.

For brands scoring 0-1/6, the highest-leverage fix is auto-generating descriptions for any page missing one and adjusting any pages outside the 80-160 band to land inside. Most modern CMSes (Webflow, WordPress with Yoast or RankMath, Shopify) have plugins that surface description-length warnings.

See also: [Open Graph Tags](/signals/open-graph-tags), [Content Depth](/signals/content-depth).