You write a meta title. You write a meta description. You save the page, check it off your list, and move on. Then a friend shares your link on Facebook and the preview card shows a different headline, or your own site's URL, or no image at all. Or you search your own page name on Google and the snippet underneath it is nothing like what you typed.
This isn't a bug in your CMS, and it isn't random. What you wrote and what gets displayed are handled by two separate systems that read your page for different reasons, at different times, using different rules. Once you know how each one actually behaves, the mismatches stop feeling mysterious.
Two Different Readers, Two Different Jobs
A search engine crawler and a social media crawler are not the same visitor wearing different hats. Google's crawler is trying to understand your whole page so it can rank it and generate a snippet that answers a search query. A social platform's crawler is doing something much narrower: it's grabbing a handful of specific tags, mainly the Open Graph set (og:title, og:description, og:image), to build a preview card fast, before someone finishes typing their post.
The Open Graph protocol, documented at ogp.me, was built specifically so any platform could pull a consistent title, description, and image from a page without having to parse the whole thing. That's why a page can have a perfectly good <meta name="description"> and still show a blank or wrong preview card on social media: if the Open Graph tags are missing, empty, or malformed, the social crawler has nothing else to fall back on. It doesn't go looking for your regular meta description as a backup in most implementations.
Google Doesn't Promise to Use What You Wrote
Your meta description is a suggestion, not an instruction. Google's own documentation on how search works, at developers.google.com/search, makes clear that snippet generation is an algorithmic process that weighs your meta description against the actual page content and the specific query someone typed. If Google decides a sentence from your body copy answers the query better than your handwritten description, it will use that sentence instead, even if your description is well within length and perfectly relevant to the page's general topic.
This is why two people searching different phrases can land on the same URL and see two completely different snippets. Neither one is "your" meta description. Both are Google's best guess at what will get a click for that specific search.
Photo by Towfiqu barbhuiya on Pexels
The Silent Failure: Tags a Crawler Never Even Sees
The most frustrating version of this problem isn't Google rewriting your text. It's a tag that never gets read at all. This happens two common ways.
The first is client-side rendering. If your Open Graph tags get injected into the page by JavaScript after the initial page load, and the crawler fetching your link doesn't execute that JavaScript, it sees an empty or default tag in the raw HTML and stops looking. Many social crawlers fetch pages quickly and don't run a full browser environment, so anything added after the fact is invisible to them, even though it looks completely normal when you view the page yourself in a browser.
The second is duplicate tags. It's easy to end up with two og:title tags in the same page head: one from your CMS template, one from a plugin or a manually pasted snippet nobody remembered to remove. Most parsers take the first matching tag they encounter and ignore the rest, so the tag that actually displays might not be the one you most recently edited. Your CMS shows you the field you last touched. It doesn't show you what else is sitting in the page head fighting for the same slot.
Why the Old Version Sometimes Sticks Around for Days
Here's a detail that trips up a lot of people: social platforms cache the Open Graph data from the first time a URL gets shared, and they don't automatically recheck it every time someone posts the link again. If you fix a broken og:image tag today, anyone who shares that exact URL tomorrow may still see the old cached card, sometimes for days, until the platform's cache expires or someone manually forces a refresh through that platform's own debugging tool.
This is a caching problem, not a "you did it wrong" problem. If your tags look right in the page source but the preview still looks stale, the fix usually isn't touching the HTML again. It's forcing that specific platform to re-scrape the URL.
The same delay shows up with search engines, just on a longer timeline. Google doesn't recrawl every page overnight, so a corrected title tag might not replace the old snippet in search results for days or weeks, depending on how often that page gets crawled. Seeing an outdated snippet after a fix isn't proof the fix failed. It's often just proof the recrawl hasn't happened yet.
Photo by Polina Zimmerman on Pexels
Character Limits Are Really Pixel Limits
The "60 characters for a title, 155 for a description" guidance floating around the internet is a rough approximation, not a hard rule enforced anywhere. Google actually truncates search result titles based on pixel width, and different letters take up different amounts of horizontal space. A title full of narrow letters like "i" and "l" can run longer than one built from wide letters like "m" and "w" before it gets cut off with an ellipsis.
That's part of why a title that looked fine in your editor's character counter shows up truncated mid-word in an actual search result. The counter is measuring the wrong unit. Writing your most important words first, so the meaning survives even if the tail gets clipped, matters more than hitting an exact character target.
"The gap between what a CMS field shows you and what actually renders is where most technical SEO problems live. I've seen teams spend weeks debugging 'ranking issues' that were really just a duplicate meta tag nobody noticed in the template." - Dennis Traina, founder of 137Foundry
Structured Data Is a Third, Separate Layer
Meta tags and Open Graph tags aren't the only things describing your page to the outside world. Structured data, usually written as JSON-LD following the vocabulary at schema.org, gives search engines an explicit, machine-readable description of what a page is: an article, a product, a recipe, a local business. Google can use this data to build rich results, like star ratings or event dates in search, entirely independent of your meta description.
The practical implication is that fixing your meta tags alone doesn't guarantee your listing improves. If your structured data is missing, outdated, or contradicts what your meta tags say, you're sending mixed signals from the same page, and search engines tend to trust the more specific, more verifiable signal over the more general one.
How to Actually Check What's Live
Since the CMS editor only shows you what you saved, not what's actually being served or read, verification means going around it.
Start by viewing the raw page source rather than the rendered page. Right-click and choose "View Page Source," or type view-source: before your URL in the address bar. This shows you the HTML a crawler actually receives, before any JavaScript runs. If your Open Graph tags aren't visible there, that's your JavaScript-rendering problem confirmed.
Next, run your markup through a structured data or HTML validator, such as the tools linked from validator.w3.org, to catch malformed tags, unclosed elements, or duplicate declarations that a casual read-through of the source won't flag.
Photo by Paul Espinoza on Pexels
Finally, use each platform's own inspection tool to force a fresh crawl instead of trusting a cached preview. These tools exist precisely because caching makes it impossible to tell, just by sharing a link yourself, whether you're seeing current data or a stale snapshot from last week.
A Short Pre-Publish Checklist
Before you consider a page's metadata finished, it's worth running through a few quick checks rather than assuming the fields you typed are the fields that will actually appear:
- View the raw page source and confirm your title, description, and Open Graph tags are present in the static HTML, not injected later by a script.
- Search the page source for
og:titleandog:descriptionand make sure each appears exactly once. - Check your title's real-world length using a tool that measures pixels or previews the SERP, not just a character counter.
- Validate any structured data so it agrees with, rather than contradicts, your meta description.
- Force a fresh crawl on any social platform where the page has been shared before, rather than trusting the last cached preview.
A resource like Moz is worth bookmarking if you want a deeper reference on how snippet truncation and SERP display actually behave across different query types, since the rules shift slightly as Google tests new result formats.
Photo by Alpha En on Pexels
Getting the Tags Right From the Start
A lot of this friction disappears if the tags are generated consistently in the first place, rather than hand-typed into a dozen different fields across a dozen different pages. That's the specific gap EvvyTools' Meta Tag Generator is built to close: it builds out the full set, title, description, and Open Graph tags, from the same source fields, so you're not maintaining two or three versions of the same information that can quietly drift apart over time.
If you're auditing an existing site rather than starting fresh, it's still worth working through the checklist above page by page. The EvvyTools tools directory has other utilities for the adjacent parts of that audit, and the EvvyTools blog covers more of the technical SEO details that don't fit neatly into a single generator tool. The full set of free calculators and generators, including this one, lives on the EvvyTools homepage if you want to browse what else is available.
The tags you wrote and the tags being read are supposed to be the same thing. When they're not, it's rarely one dramatic mistake. It's usually a small, specific, fixable gap between what your editor shows you and what a crawler actually receives, and now you know exactly where to look for it.
