Reducing the time it takes to write meta descriptions for large websites

Sometimes improving a website’s presence via search engine optimization (SEO) is very straightforward, and sometimes it’s not.

In the case of large websites, basic SEO improvements can be a massive headache. If you’ve ever optimized a large website, you know they can be prone to a myriad of SEO issues that tend to fall into one of two buckets, traditional technical SEO or issues of scale.

We could spend a lot of time talking about each, but for this article, I’d like to touch on a solution for a particular issue of scale: having to retroactively write a lot of web page meta descriptions.

I know, it’s not a sexy-sounding topic, but meta descriptions are extremely important for SEO. Along with title tags, they represent our own version of ad copy, especially since they don’t really impact query-result relevance. As long as Google doesn’t obliviate that small snippet of text, it represents our chance to capture searcher intention and influence click-through rate.

In an ideal world, the SEO practitioner in charge would act as a copywriter. Having a strong understanding of the business, audience and search intent, they would manually craft optimal, persuasive text. For small websites, this is very feasible. For larger websites with thousands of pages, this becomes an impossibility. In all likelihood, even a big business will never have enough resources to change each meta description by hand.

So what’s an SEO to do? Is the only solution to hire more writers?

Typical solutions

For some websites, especially websites where many of these pages follow the same page template, it may make sense to use the same logic and utilize templated meta descriptions as well.

Of course, this is dependent on database structure, content management system (CMS) restrictions and development resources, but it is still an excellent solution if feasible.

Have an e-commerce website with a lot of product pages? Try something along the lines of:

Buy {product name} from Store Name today. {short product description}

Is this an ideal description? Probably not, but it’s better than letting Google automatically insert random and irrelevant paragraphs of text or footer links they perceive as representative.

Is templating not an option for your company? The reality is, if you don’t provide a meta description, Google will do the heavy lifting and show a snippet anyway. Sometimes you even write a custom meta description, and Google rewrites them anyway. It may not be perfect, but the auto-generated snippets do perform well at times. Consider running a test to see.

If all of the above doesn’t really work for you, and you can’t really hire an army of copywriters, then I think you should consider another solution: semi-automation.


What if we can get machines to write meta descriptions for us? That would be pretty rad.

I remembered trying several services in the past that attempted to reduce news articles to brief summary text and figured that might be something we could leverage for meta description creation. I played around with several libraries and services and soon came to realize the results weren’t all that desirable.

Yet, they weren’t entirely useless, either.

Sometimes the text summarization algorithm from those services would input a page and pull out a very usable snippet that translated very well into a meta description or ad copy for AdWords.

Other times, what was pulled was utter nonsense, definitely no better than the cruft Google would produce if we omitted the meta description entirely and left Google to its own devices. But occasionally, when used as a basis for writing custom meta descriptions and utilized as a time-saving process, they proved to be halfway decent.

A script and a tool

Here are some scripts and tools (below the script) to help you semi-automate the description writing process.

The script

Script: Use Text Summarization Algorithms to Help Aid the Writing of Meta Descriptions 

To use the code, you’ll need an installation of Python 3 from

Next, install the sumy library using the command: pip install sumy.

Then run the script using: python

It will ask for an input text file. Input the name of the text file and supply an output filename, without a file extension. The product will be comma-separated values (CSV) with a list of uniform resource locators (URLs) and various summary descriptions produced by different text summarizations methods, several for each URL. Some of them will be better than others. Feel free to delete the worst ones or mix them together for something better.

Description tool

After that, you may need to shorten the descriptions to avoid truncation. I recommend using a title tag and meta description tool that will allow you to make changes in bulk which will make your life easier.

Search on the term “pixel width checker for page descriptions” to find a number of tools to do this or use my company’s tool on SearchWilderness:

The process of writing meta descriptions will likely be improved using machine learning techniques. The methods utilized by the code are somewhat rudimentary in a world where recurrent neural networks (RNNs) can be used to write news headlines.

So, I challenge you to improve upon this process. Make it better, and keep doing awesome technical SEO!

Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.

About The Author