Meta descriptions used to get a lot of attention in SEO. Teams wrote thousands of them by hand, page by page, because a well-written description felt like one of the few things you could fully control in a search result.
These days, much of that energy has moved to generative engine optimization (GEO), prompts and AI answers. At the same time, the number of meta descriptions many SEOs write by hand keeps shrinking, which raises a fair question about whether anyone should write them at all.
When pages loaded slowly and search results were the main way to judge a link, page titles and meta descriptions did a lot of work to help people decide where to click. Many SEOs learned early to give them a lot of weight, partly because they’re easy to understand compared to topics like JavaScript rendering.
The real question is whether anyone should write meta descriptions.
Skip to the bottom for the answer, or don’t, and keep reading.
Meta descriptions started as a simple page summary.
Have you ever wondered when meta descriptions were introduced? Well, thanks to the Baader-Meinhof phenomenon, this is guaranteed to be a question the next time you play trivia, so you’d better read this, fact-check it, and memorize it.
The meta description tag has been part of the web for nearly as long as search engines themselves. Introduced in the mid-1990s, the <meta name=”description” content=”…”> element appeared with HTML 2.0 (IETF RFC 1866, 1995) as a simple way for site owners to provide basic information about a page, including a short summary. Early search engines like ALIWEB, Infoseek, and AltaVista relied on these tags to understand what a page was about without processing every word on a page.
Once it became clear how much search engines were using meta tags, (some) SEOs did what (some) SEOs do best and started stuffing them with keywords to game the system. We all know how that ended. By the early 2000s, algorithms were already shifting from keyword counts to content quality, and meta tags lost their ability to directly influence rankings.
Meta descriptions no longer affect rankings.
Today, meta descriptions still influence how a page is represented in search results, but their role is very different from what it used to be. Google has made it clear that meta descriptions are not used as a ranking factor. In a 2020 Search Central Hangout, Google’s John Mueller explained that Google does not use meta descriptions for ranking and considers them primarily a way to influence how a page might be summarized in search results. This puts meta descriptions firmly in the category of user experience and snippet optimization rather than a direct SEO lever.
Even though they don’t impact ranking, meta descriptions still provide useful context about the overall purpose of a page. Think of it this way. Having a spoon in a drawer doesn’t make a room a kitchen. Having a spoon next to other spoons next to other utensils in a utensil organizer in a drawer also doesn’t make a room a kitchen. It could be a display at Ikea. But it does make that drawer a utensil drawer, and when you think about a utensil drawer, you think of a kitchen. When you couple that one spoon with a fridge and stove and plate, the picture becomes clearer. It’s the same way with pages on a website.
In the last few years, site owners have started reporting that Google is rewriting descriptions more frequently. Google is turning spoons into sporks, often unbeknownst to the creator of the spoon. A study by Ahrefs showed that Google rewrote about 63 percent of meta descriptions, and Google’s own documentation notes that it may choose content directly from the page when it better matches the user’s query.
In theory, the notion of pulling snippets that better align with what users are looking for sounds beneficial, and it should improve click-through rates (CTRs). In practice, we often see meta descriptions that read like Cap’n Crunch wrote them as an SEO, with Oops All Keywords.
Google is also testing a new layer of automation that goes beyond the traditional snippet rewrites we’ve been seeing over the years. In these tests, Google either fully replaces the page’s description with an AI-generated version or adds an AI-written summary of the snippet directly in the search result, marked with a small Gemini icon. Unlike past rewrites, which generally pulled text from the page itself, these new AI-generated descriptions aren’t sourced from on-page content at all.
On one hand, this could help present descriptions that are more relevant to specific segments of users. But on the other hand, it also reduces our control even further. For many site owners, the primary concern may not be with the AI generation itself (skepticism about factual accuracy notwithstanding) but with the fact that Google is continuing to tighten its grip on what is being shown within search results.
Some sites may be better off without set meta descriptions.
We’ve touched on whether you should spend time writing meta descriptions, and there’s another question worth asking, which is whether your pages should have set meta descriptions at all.
There have been a number of reports over the last few years indicating that pages rank better and have better click-through rates when site owners do not generate descriptions at all.
SearchPilot conducted a test to measure the impact on performance after removing meta descriptions, and the findings pointed toward improved results when Google created its own. After removing overlong meta descriptions on listings pages and letting Google generate the snippets, monthly organic sessions increased by an estimated 4.2 percent. This suggests that Google’s auto-generated descriptions matched user queries more effectively and helped improve click-through rate.
That Ahrefs study we referenced earlier found that even back in 2020, “25.02% of top-ranking pages don’t have a meta description.”
It seems like we have a pretty clear-cut answer. There’s reason to believe that you shouldn’t write meta descriptions because:
- The benefits aren’t worth the investment
- Google is going to replace them anyway
- Performance may improve without them
Meta descriptions do not help chatbot visibility.
Many people wonder whether meta descriptions help with large language model (LLM) visibility. We know that when a chatbot like ChatGPT doesn’t have an answer to a query, it will roughly follow this process:
- ChatGPT transforms your prompt into one or more search queries.
- Bing returns a list of search results and any available content through its API.
- ChatGPT reviews the content provided by Bing rather than crawling the open web.
- ChatGPT uses the relevant information to generate an answer and includes inline citations.
It’s reasonable to think that a meta description could factor into ChatGPT choosing its source when it searches, but meta descriptions don’t play that role.
Even if meta descriptions did play a role in being sourced when chatbots go into search mode, would that actually be worth it? As it stands, AI search and LLM sources drive a small share of traffic to many websites. While LLM traffic is growing pretty quickly, Google’s market share is still significantly higher.
If you have to write meta descriptions, you can do it efficiently.
There are going to be scenarios where removing meta descriptions from a site or ceasing to write them altogether isn’t an option. Maybe it’s built into your content management system (CMS) and you can’t change it. Maybe you’re doing work for someone and it is in your scope of work. Or maybe you’re doing it under the direction of someone who is a believer in the old school ways. Whatever the reason, there are things you can do to ensure you’re not wasting too much time, especially if you’re managing a site with hundreds or thousands or billions of pages.
You can use ScreamingFrog and OpenAI to generate meta descriptions in bulk.
- Open Crawl Configuration in ScreamingFrog > Spider > Extraction. Select Store HTML.
- While still in the Crawl Config, open the API (application programming interface) Access menu and go to OpenAI.
- Input your API Key and select connect.
- Note: You have to pay to use the OpenAI API, but the cost versus time-savings is negligible. In order for this to work, you need to add credit to your OpenAI balance or else you’ll receive a “429 – Too Many Requests” error when you try to run your crawl with the API.
- Go to Prompt Configuration and select +Add (not +Add from Library).
- You can rename the first field “Generate Meta Descriptions” if you’re going to be running multiple prompts.
- Click the Gear icon ⚙️ in the Enter Prompt field and enter your prompt.
- I pay a prompt engineer a 6-figure salary to come up with this:
- Select OK and you’re good to go.
- If you’re still seeing a red exclamation mark on the prompt screen, it likely means you haven’t selected to store HTML yet.
- Don’t forget to configure the rest of your crawl. Add your sitemap link, connect to Search Console, etc.
- As the crawl is running, head to the AI tab and watch your responses fill up. If you’re not satisfied with the output you’re seeing, stop your crawl and tweak your prompt.
- Once it is complete and you’re satisfied, export your file.
Now you’ve possibly generated hundreds to thousands of meta descriptions in minutes. In the past, writing that many meta descriptions could have taken multiple weeks’ worth of time and cost thousands of dollars to attribute to $0 in sales. Now, it can take one hour and cost ones of dollars to attribute to $0 in sales.
At this point, you may be wondering if implementing all of those meta descriptions on your site is as easy as generating them. It isn’t, but that’s another article and another decision you’ll have to make. If you want help sorting out which on-page work is worth your time, our SEO team can help.
Meta descriptions matter far less than they used to.
- Meta descriptions no longer impact rankings, and Google frequently rewrites them based on page content or AI-generated snippets.
- Removing meta descriptions can improve click-through rate because Google often produces query-matching snippets more effectively on its own.
- Google is now testing Gemini-generated descriptions that are not tied to on-page content, further reducing publisher control.
- Meta descriptions do not influence LLM visibility because chatbots rely on search API responses, not metadata.
- Writing meta descriptions is often unnecessary, but some projects still require them due to specific expectations or workflow limitations.
- ScreamingFrog and OpenAI make it possible to generate large volumes of meta descriptions quickly and efficiently.
- The primary value of meta descriptions today is limited to user-facing snippet clarity rather than direct SEO performance.