Forget what you know about title tags

Google threw a few coins into the machine learning arcade and started rewriting meta-titles around August 23rd. The tech doesn’t seem to get it right every time, yet. Some webmasters reported backwards spelled brands and some government authorities turned into tax avoidance schemes (just google “hrmc”). Let’s say the SEO and webmaster community wasn’t thrilled.

I started a poll on Twitter and got very mixed results. I forgot to add a “nothing changed” option, though, so take this with a grain of salt.

It’s too early for recommendations. We still need to learn a lot about this change, but I thought it was a good occasion to revise meta titles - one of the most impactful ranking signals in my mind - and some of the common myths around them.

The point is not to pretend I know what a perfect title is. The point is that there is no perfect title - only iteration. The key to success here is testing, testing, testing.

Title tag studies and experiments

You think titles have to be 60 characters long? Think again! That’s actually a length the community came up with.

To quote Barry Schwartz here: “The display of a snippet is not directly related to what Google uses for ranking purposes.”

Just in February of 2021, Gary Illyes confirmed that the displayed length of a title is not equivalent to what Google uses for ranking purposes (source).

To test that, Charles from Pages set up a title test (source) with 36 pages that have various title lengths. In one case, the title tag had 1,000 characters before the keyword even appeared!

Google ranked all pages for the test keyword and even tried to rank the pages for all the keywords that appear in the title tag! However, the page with just the keyword in the title ranked best.

What does that say about the importance of the title carrying the keyword?

Several ranking factor studies have devalued keywords in the title as a ranking signal over the last years (source). In my mind, that doesn’t mean you should intentionally leave it out but also not over obsess about it.

Especially the position of the keyword doesn’t seem important anymore. Page optimizer set up a test (source) that resulted in better rankings when the keyword sits at the end of the title. Mind you, it was a small test.

Marie Haynes tested different variations of keyword positions in the title tag (source) and found varying results. Most interestingly, Google rewrote some title tags with unrelated content or too long ones - and that was in 2016!

Having the brand in the title seems to be important in some cases, though. A search pilot experiment showed a +15% increase when the brand stood at the beginning of the title (source).

Alluding to the value users are looking for seems to have a big impact on organic traffic. Rankscience tested adding (example) to titles of (source) of a developer community and found a 14.8% uplift in organic traffic. At the same time, taking numbers out of listicle titles when users expect them can harm organic traffic. Searchpilot found a -16% negative impact on organic traffic (source). For relevant topics, adding the year can bring a +5% uplift and more (source).

Aggregator vs. Integrator titles

I hope these case studies and experiments show that title tags are context-dependent. Or, in plain terms, search intent comes with an expectation and your title better hits or tops it!

Title optimization is also a different game for aggregators vs integrators. Aggregators like marketplaces, listing sites, or e-commerce stores have many pages on the same template, which means they can run simple split or multivariate tests. A good example is Etsy (source).

Integrators like SaaS companies or publishers, though, need to test titles with much smaller sample sizes and have a harder time finding pages for control groups.