Google rewrites your meta title tags 76% of the time. Which kind of begs the question: should we even bother writing them at all?
There’s a great debate going on in the SEO industry right now about whether metadata, meta titles and meta descriptions, in particular, are still worth your time. Are they still relevant, or are you polishing a brass doorknob on a house that’s about to be demolished?
At Excite Media, metadata is one of the first things we look at when we start an SEO package with a new client, and yes, that includes writing and updating meta descriptions on every page we optimise. Although what we write may not always survive to the search results exactly as written, keeping it organised and readable for search engines and AI systems like ChatGPT, Claude, and Gemini still definitely matters.
So if you’re wondering what everyone in the industry is arguing about and whether there’s more to metadata than just your title and description tags, this article is for you.
- Google rewrites titles roughly 76% of the time, but your original wording still influences the final result.
- Structural metadata like schema, Open Graph, and URL slugs is becoming more important than traditional titles and descriptions.
- Indexation basics like robots.txt and crawl access are the real non-negotiable, since nothing else matters if Google can't reach the page.
- AI-referred traffic is growing fast and converting significantly better, changing the goal from chasing clicks to earning citations.
- Metadata isn't dead, it's just evolving, the winners in 2026 are the ones adapting their focus rather than clinging to old habits.
- Google not always keeping our wording doesn't mean we skip writing meta descriptions, we write and maintain one for every page.
What is metadata?
- Meta titles
- Meta descriptions
- Meta tags
- Structured data
- Alt text
- Canonical tags
- Hreflang
- Sitemap
- Robots.txt
- Open graph
We’re mostly focusing on meta titles and meta descriptions here, but we’ll also touch on the rest of the list and whether their importance is growing or fading.
How we write and update meta titles and meta descriptions
How you actually do this comes down to whatever CMS you’re working with.
On WordPress, it’s usually a plugin doing the work, most commonly Yoast or RankMath. All you have to do is type your title and description into the fields, hit update, and that’s it.
Webflow makes it even easier, with the same kind of fields built straight into page settings, so there’s no plugin to install in the first place. Whatever platform you’re on, chances are there’s a similarly simple way to get it done.
The one thing worth remembering is to keep titles under 60 characters and descriptions under 160. Go over, and Google will happily chop them off mid-sentence.
Google’s use of metadata
As mentioned in the opening line of this article, Google changes what we’ve submitted as our meta titles and descriptions and by quite a lot, actually! Some reports suggest this can happen as often as 75% of the time, meaning the metadata you write is not always what ends up appearing in search results.
Google is pretty transparent about this as well. On their developers’ website, they say:
Google’s generation of title links on the Google Search results page is completely automated and takes into account both the content of a page and references to it that appear on the web. The goal of the title link is to best represent and describe each result.”
Essentially, what they’re saying is: tell them as clearly and accurately what your page is about, and they’ll organise it and present it how they believe to be most useful to Google users.
Google’s official position
The biggest clue as to how Google feels about meta descriptions sits in their Control Your Snippets page.
It’s pretty clear in this document that quality is important, but meta descriptions are treated more like editorial guidance than a ranking factor.
As mentioned earlier, one reason SEOs spend time on meta descriptions is to improve click-through rate. Google even describes them as being “like a pitch that convinces the user that the page is exactly what they’re looking for.” The tricky part is that if Google rewrites snippets so often, it becomes harder to treat them as a reliable CTR lever.
For larger database-driven sites, such as product aggregators, writing every description by hand is often unrealistic. In those cases, programmatically generated descriptions can be a good option, as long as they are still human-readable, useful, and varied from page to page.
Title Tags: Google's Clear Position
Google uses a number of different sources to determine the title tag it will use in its results, not just what you set as a title tag.
It draws on the title, H1, site schema, prominent text, and more to land on something accurate, and it will override anything outdated, inaccurate, keyword-stuffed, or boilerplate. That part’s automatic, and there’s no opting out.
What is clear is that these title tags can be really important in user decision-making. A clear, relevant title can help someone quickly understand what the page offers and decide whether it is worth clicking through.
Structured Data: Google's Strong Recommendation
Structured data markup is a form of metadata that provides information about webpages in a standard, easy-to-read format for search engines.
Google provides four case studies on its introduction to structured data page, suggesting that it values this highly.
It prefers J-SON format, and again prefers quality over quantity, citing “it is more important to supply fewer but complete and accurate recommended properties rather than trying to provide every possible recommended property.”
There’s also a theory going around that structured data helps with AI search rankings. Google pushes back a little here, saying it’s not required for generative AI search and there’s no special schema you need to add, though they still recommend it as part of a broader SEO strategy since it helps you qualify for rich results.
So, take from that what you will. It may not be a direct AI search ranking lever, but clean, accurate structured data is still worth doing properly.
Should we use metadata or not?
Now, let’s look at the key arguments for focusing on metadata and then for putting more time into other areas of SEO.
As is often the case when speaking to an SEO professional, the answer to much of this is “it depends”, so we’ll also take a look at the middle ground as well.
The case for metadata
Fair to say that the argument for pouring hours into traditional meta titles and descriptions is weaker than it used to be, but there are plenty of areas of metadata that are not just important, but essential for SEO.
Title tags and descriptions still influence click-through rate
Even when Google rewrites your title, it doesn’t eliminate the ranking benefit. Your title still gives context to the page and usually carries your most important keywords.
Well-crafted titles increase CTR when they do appear (which is more often than meta descriptions). Even if Google rewrites it, your words have still influenced what appears in those title tags.
Meta descriptions, although not always published exactly as you’ve written them, can still improve CTR by up to 5.8%.
Open Graph metadata shapes how AI systems represent your content
Open Graph tags ensure correct representation outside traditional search; these can include AI previews, summaries and content surfaces.
In 2026, as generative AI systems increasingly pull structured metadata to understand web content, having correct Open Graph metadata is also a signal that your content is well-organised and trustworthy.
As AI systems pull structured data, Open Graph metadata signals trustworthiness to those systems.
They’re also useful for when someone shares a link – your Open Graph metadata will display when the link information expands.
Indexation is the real non-negotiable
Before metadata can influence how a page appears in search, the page needs to be accessible, crawlable, and indexable.
Without tools like robots.txt, indexation controls, and crawl directives, it becomes much harder for search engines to understand which pages they should access, crawl, and include in their results. In some cases, poor technical control can prevent important pages from ranking at all.
While this isn’t directly related to your meta titles and descriptions, this is a fundamental part of your SEO foundations.
Schema Markup has measurable ROI and helps AI systems understand and cite your content
Google’s own case studies, which we cited earlier, demonstrate 25-82% CTR uplifts with well-structured data.
Although increases like these may not be the case for every website, Google is clearly saying that structured data, done right, is an essential component of a website hoping to rank in its search engine.
Multiple sources have also indicated that schema gives search engines and AI systems the context they can’t safely assume on their own, clarifying entities, relationships, and meanings, the raw ingredients behind AI Overviews and entity-based search.
This is particularly important as AI models rely on entity graphs to determine which sources to reference.
URL slugs are being read by LLMs for relevance
Your URL structure is a technical piece of metadata, and well-built URLs are becoming increasingly important, especially for showing up in AI features.
Specific slugs are tending to outperform generic ones. For example, in our blog ‘Thinking of building your website with AI? Here’s what you need to know’, we’ve optimised the slug to /building-website-ai-what-to-know/ – a very specific description of the webpage.
The case against metadata
The case against metadata is mostly about time, not usefulness. We’ve already shown some metadata really matters, so why bother putting effort elsewhere?
Google rewrites meta descriptions most of the time
The obvious argument… If Google is using multiple sources to produce its own meta description, why spend your time writing them?
Let’s look at some examples we shared earlier and do the math. A well-crafted description can lift CTR by 5.8%, and if Google rewrites 76% of them (keeping only 24% of what you wrote), that 5.8% shrinks to roughly a 1.4% actual impact.
Which begs the question, is the time you put into writing your meta descriptions worth a 1.4% increase in CTR?
That’s a fair question to sit with, but it’s not one we’ve answered by dropping meta descriptions from our process. We still write and update one for every client page we work on. It’s cheap insurance against Google having nothing decent to work with; it helps with indexing new pages, and on the pages where Google does keep our wording, that 1.4% adds up across a whole site.
Time spent on descriptions may have a better ROI elsewhere
US-based agency Seer Interactive produced a case study where they redirected time spent on metadata towards working on content optimisations and reported an increase in organic traffic for an e-commerce client.
However, they did suggest that meta descriptions were a useful tool for helping search engines to index new pages. They cautioned that these may not be worth rewriting in the future, instead using that time to optimise the actual text on the page.
That’s roughly where we land too: we still write a meta description for every page, we just don’t treat rewriting an existing, well-performing one as a priority task once a page is established. New pages and launches always get one from day one.
Zero-Click search makes traditional metadata less impactful
AI Overviews have changed how Google users interact with search results. Users are less likely to click a blue link if there’s an AI Overview and more likely to search again if they don’t get the answer they want.
If meta descriptions exist to drive CTR, and people aren’t clicking through anyway, what’s the point in agonising over one Google’s just going to rewrite?
Research from Spark Toro shows that AI answers draw directly on answers in the top 10 search results and lead to clicks in only ⅓ of searches, suggesting that optimising for CTR could be redundant.
Meta keywords are completely dead
Google had previously used keywords in metadata (not to be confused with keywords in your title) as a ranking factor, until they confirmed back in 2009 that it had zero impact. It’s been dead a long time, so let it rest.
Schema's direct effect on LLMs is not fully confirmed
We mentioned schema and its importance in the section arguing for metadata.
It is worth noting that Google has clarified that AI features don’t strictly require schema and that LLMs don’t crawl schema markup directly. Tests by experts such as Mark Williams-Cook have argued against schema’s role in LLM features, claiming that they prefer natural language models.
Google also deprecated seven schema types in late 2025, including Practice Problem, Dataset for general search, Sitelinks Search Box, SpecialAnnouncement and Q&A markup.
Which raises a fair question: if Google’s trimming schema types, how much should you really be pouring into it long term?
The Middle Ground
The answer, of course, lies somewhere in between the two arguments. But we feel it looks like this.
Traditional vs. Structural Metadata
- Traditional metadata (title tags, meta descriptions) = diminishing returns, Google frequently overrides
- Structural metadata (schema, robots tags, Open Graph, URL slugs) = still matters, especially for entity clarity and AI citation
AI-Referred traffic is growing exponentially AND reportedly converting better
We’re still in the early days of the AI journey, with developments in technology happening practically every week.
Research from Frase shows that AI sessions grew by more than 5x year-over-year between mid-2025 and now, and that visitors from AI convert at 4.4x the rate of standard organic visitors. Not only that, they spend 68% longer on websites.
If that is the case, there is a real desire to move from features to clicks, which means trying to be cited more often on AI.
That will make metadata, especially structured data and entity signals, more important, not less.
The Real Metadata Hierarchy for 2026
We’ve mentioned a lot of different metadata in this article. Which ones are still working, and which ones should you skip?
Must-do (non-negotiable):
- robots meta tags / noindex control (access to indexation itself)
- 'title' tags (Google still reads and uses these)
- URL structure (slugs matter)
Should-do (high ROI when done right):
- Schema markup for rich results (25–82% CTR uplift documented)
- dateModified in Article schema (AI recency signals)
- Open Graph for accurate representation
- URL-based optimisation
- Meta descriptions for every page (we write and maintain these as standard, even though Google won't always keep our wording)
Selective (context-dependent):
- Rewriting meta descriptions on long-established, already-indexed pages (lower priority than getting a first one live)
- Full-page schema implementation (focus on high-value schema types first)
Skip (genuinely waste of time):
- Meta keywords
- Over-optimised title tag keyword stuffing
- Manual descriptions on every page of a 50,000+ URL site
