Three pieces of meta-tag advice are repeated everywhere and are now wrong in ways that change what you should write.
Meta description length
The commonly quoted limit is a character count, and the actual limit is a pixel width. That distinction matters because characters are not equal widths — a description full of capitals and wide letters truncates earlier than one of the same length in narrow lowercase.
Desktop results cut at roughly 920 pixels, which works out at about 155 to 160 characters of typical prose. Mobile is considerably tighter, around 120.
| Surface | Approximate limit | Safe target |
|---|---|---|
| Desktop | ~920px, ≈155–160 chars | 150 chars |
| Mobile | ≈120 chars | 120 chars |
| With a date prefix | Date consumes ~10 chars | Allow for it |
Write the essential point in the first 120 characters so it survives on mobile, and treat everything after that as a bonus that desktop users may see. Descriptions that trail off mid-sentence at the cut are a small, avoidable signal of carelessness.
Google will probably rewrite it anyway
This is the part most advice omits. Google frequently replaces the supplied meta description with text pulled from the page, when it judges that text a better answer to the specific query.
Studies of large result sets consistently find rewrites on the majority of results — the exact proportion varies by study and by query type, but it is common enough that a description should be treated as a proposal rather than a specification.
The practical consequence is that on-page copy matters more than the meta tag. If a page answers a question in a clear sentence near the top, that sentence is what Google has to work with when it rewrites. A page whose body text is vague gives it nothing better than the description you wrote, which sounds like an argument for the description and is really an argument for the body text.
Keyword density is not a thing any more
There is no target percentage. There has not been one for a long time, and the tools that report an ideal range of two or three per cent are reporting a convention rather than a measurement of anything Google does.
Search engines moved to embedding-based understanding: a page is matched on meaning, and related terms, synonyms and context all contribute. Repeating an exact phrase to hit a ratio adds nothing, and past a point it triggers the spam policy on keyword stuffing.
- Density has no target. Write naturally and stop counting.
- Stuffing is a policy violation, not merely ineffective.
- Related terms help — they signal genuine coverage of a subject.
- It is still useful as a diagnostic, in one direction only.
That last point is the reason a density checker still earns its place. Running one over a draft occasionally reveals that a phrase appears thirty times because it was worked in deliberately, and that page reads badly to a human. Keyword Density Checker is worth using as an over-optimisation alarm, never as a target to hit.
Open Graph, tag by tag
| Tag | Controls | Required? |
|---|---|---|
| og:title | Headline on the share card | Yes |
| og:description | Text below the headline | Effectively |
| og:image | The picture — the whole card, visually | Yes |
| og:url | Canonical URL for the share | Yes |
| og:type | website, article, video… | Yes |
| og:site_name | Your site name on the card | No |
| og:image:alt | Alt text for the image | No, but do it |
The image does most of the work. A share card without one renders as a small text link that is passed over, and with one it occupies several times the space in a feed. Everything else on that list is refinement by comparison.
Getting the image right
- 1200 × 630 pixels, a 1.91:1 ratio. This is the size every major platform crops to.
- Under 8 MB, and ideally well under — large images time out during fetching.
- Absolute URL, including the protocol. Relative paths fail on most platforms.
- Keep text away from the edges. Some surfaces crop to a square.
- Twitter falls back to Open Graph if no twitter: tags are present, so you rarely need both. Add
twitter:cardset tosummary_large_imageand stop there.
Facebook and LinkedIn both cache aggressively. After changing an image, the old one keeps appearing until you force a refresh through the platform's own sharing debugger — which is also the fastest way to see what the card will actually look like before posting it anywhere.
What to spend effort on
Ranked by return, the order is not the one most checklists imply.
The title tag is the only one of these that meaningfully affects ranking and it strongly affects clicks. The Open Graph image affects whether anything shared gets clicked at all. The meta description affects clicks where Google keeps it. Keyword density affects nothing.
Meta Tag Generator produces the full set including Open Graph and Twitter, which mainly saves you from the two errors that break cards in practice: a relative image URL, and a missing og:url.
Frequently asked questions
How long should a meta description be?
Around 150 characters, with the essential point in the first 120 so it survives mobile truncation. The real limit is pixel width — roughly 920 pixels on desktop — so a description in wide capitals truncates earlier than the same character count in lowercase.
Does Google use my meta description?
Often not. It frequently rewrites the description using text from the page when that better matches the query, and studies find rewrites on the majority of results. Clear answer-shaped sentences near the top of your page give it better material to work with.
What is the ideal keyword density?
There is not one. Search engines match on meaning rather than term frequency, and the two-to-three per cent figure quoted by older tools reflects nothing Google does. Keyword stuffing is a spam policy violation, not just ineffective.
Is a keyword density checker still useful?
As an alarm, not a target. Running one over a draft can reveal a phrase repeated thirty times because someone worked it in deliberately — which reads badly to humans and risks the stuffing policy.
What size should an Open Graph image be?
1200 × 630 pixels, a 1.91:1 ratio, under 8 MB, referenced by an absolute URL including the protocol. Keep important text away from the edges, since some surfaces crop to a square.
Do I need Twitter card tags as well as Open Graph?
Rarely. Twitter falls back to Open Graph tags when no twitter: equivalents are present. Adding twitter:card set to summary_large_image is usually all that is worth duplicating.
Why is the old image still showing when I share a link?
Facebook and LinkedIn cache Open Graph data aggressively. Force a refresh through the platform's sharing debugger, which also previews the card before you post it anywhere.