The humble title tag. Probably the single most important 50-60 characters of that piece of content you’ve just written.
Perhaps you’ve found this post because you’ve spent hours pouring your soul into a piece of writing and now you’ve realised people will only read it if you write a good 50-60 characters. Or maybe it’s just that your boss told you that he needs quick wins for your product pages and so you’re turning in desperation to the ol’ title tag. Writing a good title tag is part art, part science. How do you do it?
The title tag of a page is the HTML tag which is used to summarise the content of your webpage. It’ll be used by search engines as the title in the search engine results:
In your browser tab:
And even as a fallback in social sharing posts.
It isn’t the same thing as the on-page title! An on-page title could be written as a variation of your title tag, or something completely different. If we take a look at the article I’m using as an example we can see that the brand isn’t on the on-page title.
If you want a more severe example take a look at this Redbull article.
A title tag should typically be 50-60 characters. Technically Google’s maximum size is 600px. This usually works out at about 50-60 characters.
And if we just do one, you usually don’t get the best results. For example, using the title from the blog post above:
We want to maximise how clicky our titles are without… you know… lying, mentioning that one trick dentists hate and crucially without compromising on summarising the page.
The title is primarily for people arriving on your site from Google. We’re not trying to pull people in who are idling. Those people are on Facebook, TikTok, Youtube, Instagram etc. (I know we did mention above that the title can sometimes be for social, but you can overwrite that if you’d like!)
The audience for your title is someone searching with an intent & that always comes first.
The process is quite different now depending on if you’re writing for a single article, or a template.
Write the article. It’s far easier to write a title when you know what you’ve written about. (This is assuming you know what you’re writing about, otherwise, sometimes headline writing can be a good way to generate ideas.)
Pull out the primary purpose/point of the article. No clickiness yet, just the factual summary.
Try to summarise what someone might search to find your article. Aim for the simplest most basic version of it. Search that term, take the top 5-10 articles which rank for it, plug them into a tool like Ahrefs, SEMRush, Searchmetrics, Brightedge etc. and download all the keywords those articles rank for.
If the top 5-10 articles look nothing like yours either:
Once you’re happy with the phrase, take that big list of keywords and look for any other commonly occurring phrases you’re missing and take note.
We’re going to continue using my old article on log analysis as an example. Because it doesn’t have a great title…
First search phrase pick: “log analysis”
If we look up this keyword these are the top articles (only 3 shown below). Clearly we can see here that none of these articles are about search log analysis, I probably need to change my keyword:
Second search phrase pick: “seo log analysis”
Yep, that search result looks far better. We’ve still got a short phrase, but now the articles are now on topic with my own:
And then get the most common keywords from that list. This ngrams tool is a nice way to do it. We get:
If we pull out the big generic words which would also apply to my article we get:
And possibly also:
Now we’ve got all the factual words we’ll want in our title and brand.
What inspiration can we get for the clicky part? Lets quickly blast through a couple:
Then we try to write as many headlines as we can, but without trading away our relevance and factual keywords.
Back to our previous example.
We’ve got our important factual words. We also know we want SEO as without that the intent of results shown wasn’t correct. Together those 4 words (without server) take up 18 characters. Which gives us roughly 32 characters left to play with. Let’s also look at our current title and see what we’re working with:
We can see I’ve used “Complete Guide” to try and make it clicky and that I’ve also put the method of analysis “BigQuery” into the title. Both of these we could definitely play around with. Now we just try to write as many titles as we can.
I started with the restrictions and gradually just ignored them in my attempt to get to 20 titles. I didn’t get there. Sorry Hannah.
How do we decide which is best?
Honestly, it’s savagely hard to pick the right title by yourself. Of all the title tag tests we’ve run, only one in five is typically positive. When I first started in search, I thought titles were the easy win. About a year and a half of running endless title tag split tests and I’m no longer convinced.
If you can test it. The two easiest ways for a single article are:
The above process works great if all you need to write is a single title.
But if you’ve got a template with hundreds of thousands of pages, then you can’t really do that. Well, you could, but it would be exhausting. Instead, we’re going to need a format for a title that we can apply to all our pages, to make our template shine. That previous process won’t cut it.
We’re going to start by trying to summarise the attributes of the page in as much detail as possible. This will give us an idea of what pieces of detail we can pull into our titles across our template.
I’ve pulled two page templates from rightmove.co.uk (this isn’t every page template but we’re keeping it simple):
Our templated page matches a specific intent. We need to figure out how to represent that in a title tag.
Two things make this hard:
We need to try and make a title which:
If we’re really struggling perhaps these pages shouldn’t even exist. But that’s a conversation for another day.
We have two templates:
In this case, it’s pretty simple. For sale & to rent are clearly the important keywords we need to keep each template different. We can see that by looking at the SERPs. Changing those keywords, changes the results from for sale to rent.
Within our template, we have lots of different locations.
In order to keep the pages in our template different, we’re going to need the location in the title.
But anytime you work with titles it’s going to get messy.
Take our previous example. Rightmove actually has pages for Manchester & Greater Manchester. One ranks for properties and the other for flats. Something is clearly going on there. Uh oh.
Should that change what we do?
When we’re working at scale, patterns are going to breakdown. There hopefully is an underlying pattern, but look long enough and you’ll find exceptions. All we can do is do our best. Make a reasonable guess at what is going on and spoiler for stage 6. Test.
This is exactly the same as step 3 for articles.
To keep it brief, we’re going to just stick with the properties for sale template for the rest of these steps! Running this example with the top phrases for “properties for sale in manchester” we get:
Words to note here are all fairly self-explanatory:
We know what we need to include to make the intent of our page clear.
Now let’s use that as a base and write as many titles as possible.
We want to:
A general difference between this and individual articles: If you end up with an entirely factual template title that is far more acceptable here than with an individual article.
Let’s have a go at writing titles for our category pages
Our base is:
Let’s make variants:
That’s a lot of variations. We even managed to fit in their tag line at the end.
Just like with articles we’re going to end up with a list of titles and unsure which one will be best. Far more than with individual title tags, it’s really really important to split test.
If you can test at all I’d highly recommend it.
If you can’t test, you can at least lean on our tests, I’ve got results from those in the next section.
We’re lucky enough to have access to SEO split testing software. It lets us test different titles & accurately measure the impact on organic traffic. We’re about to talk about the different results we’ve learned, so it’s important to briefly talk about the assumptions implicit in these results.
You can only run SEO split tests on large groups of similar pages (e.g. all category pages, all listing pages etc.) and that means our results are from certain types of websites:
I think you can learn a huge amount from these tests, but it’s still important to bear those assumptions in mind.
Writing titles is really hard. We mentioned this above, but let’s look at our numbers in slightly more detail. We’ve run many title tag tests across different industries. Our results break down as follows:
Oof. 78% of the time title tag tests fall flat or actually harm the website. That makes testing super important. It’s not impossible you could work on a website where you never have a positive title tag test. Nothing you try will ever work. Without testing, you’d probably still roll out those titles. Just spotting the failures and not rolling them out will save you a huge amount of traffic.
With a single article, this isn’t so worrying, you’ve got a far larger creative space to play in and if it does go wrong, it’s a far smaller proportion of your traffic.
If you’re changing titles on big page templates, please make sure you test them!
Broadly most title tag tests have an impact between 4-15% in either direction.
You can see a distribution of our title tag tests below.
Most title tag changes are unique to a website, changing words and phrases which don’t generalise well from website to website. However, there are some more common patterns we’ve been able to test.
50% of our title tag tests involving adding the price into the title have been positive. Not only do we get to put a number into the title, but it also provides more information.
One of our consultants thinks this is down to whether or not Google can find the price you put in the title on the rest of your page – i.e. are you being honest about price. We also think it may make a difference depending on how competitive you are on price.
We haven’t had the chance to test this a huge number of times, but so far this change has been positive in the niches where we’ve done it. The shameless putting 2019, 2020 in the title has helped.
When you have lots of automatically generated titles, it’s common to end up with titles that are too long.
We’ve run a number of tests about shortening these titles and nearly all of them have been null (~80%). They’ve also never been positive. Our best current theory is that the templates which often end up with long title tags are typically attracting long tail traffic. When they are truncated, they’re still the only relevant result and so continue to rank, perhaps for long tail queries, keyword stuffing isn’t a problem.
Having said that I’d still say it’s worth trying to shorten your titles. If you manage to cut 4-5 characters from your title with no effect, you could use that space to add price or something else which may have an effect.
We’ve run several tests to put emojis into title tags and so far it hasn’t helped. Sorry folks 🙁
I mean c’mon. Marketers can barely be trusted with FAQ schema, can you imagine what we’d do to Emojis.
We’ve tried some title tags for category/listing pages which were very different, actively calling out to the user in the SERPs.
These did not work.
We tested using localised versions of phrases. This wasn’t single letter changes (like s for z in UK vs US), but entire words e.g. pants instead of trousers.
This was notably positive (~20-25%).
We’ve seen mixed results from this. We ran a split test & found removing “online” from title tags had no effect on one particular client. Outside of our split-testing platform for a different client, we removed the word “online” from the title of an online store.
Our rankings for the terms including “online”, dropped and we quickly put it back in.
We usually see the impact of a title tag in 3-5 days. We’ve had a couple which has taken longer, but this is the majority. The previous caveats are of course important here, we typically work on larger websites, which are heavily crawled.