Title Tag Rewrites: 7 Months Later

  • Post author:
  • Post category:SEO

Back in August, we analyzed 10,000 SERPs and found that Google was rewriting 58% of the title tags we were able to track. In September, after some serious objections from the SEO community, Google released the following statement:

Based on your feedback, we made changes to our system which means that title elements are now used around 87% of the time, rather than around 80% before.

This immediately raises two questions. First, has the situation improved? Second, why the huge mismatch between our numbers (and similar numbers by others in the community)?

Rewrites by the numbers

We collected new data on March 2, 2022 from the MozCast 10,000-keyword tracking set. Here are the basic stats, which are very similar to what we found in August 2021:

  • 84,639 page-one results

  • 71,265 unique URLs

  • 57,832 <title> tags

  • 33,733 rewrites

So, letโ€™s compare the August 2021 rewrites to the March 2022 rewrites:

Technically, the numbers did go down, but this probably isnโ€™t the news you had hoped to hear. If 57% of titles in our study were rewritten, then โ€” I think we can all agree with this math โ€” 43% did not get rewritten. So, how do we reconcile our 43% with Googleโ€™s 87%?

Truncation, from simple to โ€ฆ

First off, our definition of โ€œrewriteโ€ is extremely broad, and it covers truncation, where Google just runs out of physical space. In August, I took a pretty simplistic view of truncation, but letโ€™s try to give Google some benefit of the doubt. Iโ€™m going to dig into three forms of truncation, starting with the simplest:

1) Simple truncation

The simplest form of truncation is when Google cuts off a long title but preserves the original text from the beginning. For example:

No one is doing anything wrong here โ€” the IRSโ€™s <title> is accurate and descriptive, but Google ran out of space. They didnโ€™t take any liberties with the text.

2) Midstream truncation

Letโ€™s review another form of truncation, with this example from the Linksys website:

Again, Google truncated a long <title>, but here they removed the branded text from the beginning and started with the more unique, descriptive text. Is this a rewrite? Technically, yes, but itโ€™s a direct excerpt and the โ€œ…โ€ clearly implies truncation to searchers.

3) Excerpt truncation

Finally, we have situations where Google uses a portion of the <title> tag, but they donโ€™t clearly indicate truncation with an ellipsis (โ€œ…โ€). Hereโ€™s an example from Congress.gov, a site Google is unlikely to view as spammy or in need of editorial revisions:

I donโ€™t think Googleโ€™s trying to hide the truncation here by removing the ellipsis โ€” the truncated title is a complete thought/phrase within the original title. In some cases, is this the excerpt the creator would have chosen? Maybe not, but I would still generally call this truncation.

All told, these three forms of truncation accounted for almost exactly one-third of the โ€œrewritesโ€ that we observed. These forms were distinct enough that we could separate them. From here on out, it gets a bit more complicated.

Title additions (brand & local)

In addition to truncating long titles, Google sometimes adds information they deem relevant to the end of a display title. The most common addition is โ€œbrandโ€ information (using the term loosely) that wasnโ€™t present in the original <title> tag. For example:

I kind of love this title, and you should definitely ride Top Thrill Dragster at Cedar Point if youโ€™re a coaster fan, but notice here how Google has appended โ€œTouring Ohioโ€ to the end of the display title. This kind of add-on is very common, occurring in almost 14% of our observed rewrites.

In some cases, adding the brand text caused Google to truncate the title prior to the addition. See this example from Goodreads โ€ฆ

While the rewrite here is intended to be beneficial, this can cause problems with long brand names. Anecdotally, though, Google seems to be doing a better job of this in the past few months, and most brand identifiers are of reasonable length.

Finally, in a few cases, Google appended location information. For example:

Itโ€™s not clear what situations trigger this added location information, but it does show that Google is considering appending other forms of relevant information that could drive future rewrites and go beyond brand tagging.

Capital-R Rewrite examples

We can argue about whether truncation and addition are real, Capital-R Rewrites, so how about the situations where Google is clearly making substantial changes? Some of these situations โ€” even working with a moderately-sized data set โ€” are hard to classify, but Iโ€™ll cover some major categories.

1) Maximum verbosity

I almost said โ€œkeyword stuffing,โ€ but thatโ€™s a judgment call and isnโ€™t always fair in these cases. Granted, there are legitimate cases of keyword stuffing, like this example:

Prior to August 2021, Google mightโ€™ve just truncated this title, but now theyโ€™re saying โ€œYeah, noโ€ and replacing the entire mess. Other cases arenโ€™t so clear, though. Consider this one:

AMC hasnโ€™t really done anything spammy here โ€” this <title> tag is likely a direct reflection of their site architecture. In this case, though, Google has gone beyond truncation and rewritten the title, including replacing pipes with hyphens, removing โ€œMovie Timesโ€ (which is arguably redundant with โ€œShowtimesโ€) and pushing the site/brand up.

2) Minimum verbosity

Some people have too much to say, and some people are too quiet (Iโ€™m afraid I know which side I fall on). Hereโ€™s a case where the title didnโ€™t quite provide enough information:

In many of these cases, like displaying just the brand name, a generic placeholder like โ€œHomeโ€, or โ€“ in one notable case โ€“ a code placeholder (โ€œtitleโ€), itโ€™s likely the culprit is an overzealous CMS default setting. These are clearly Capital-R Rewrites, but I would argue that Google is generally adding value in these situations by rewriting.

3) Excessive superlatives

Sometimes, we marketers get a little carried away with colorful language (in this case, the family-friendly kind). Google still seems to be disproportionately rewriting <title> tags with certain superlatives, even when they may not seem excessive. Take this example:

This is a case where Google replaced the <title> with the contents of an <h1> โ€” while itโ€™s not a bad rewrite, it does feel aggressive to me. Itโ€™s hard to see how โ€œ21 Best Brunch Recipesโ€ is wildly over the top or how โ€œ21 Easy Brunch Recipesโ€ is a major improvement.

4) Miscellaneous nonsense

Itโ€™s hard to measure the real head-scratchers, but anecdotally, it does appear that Googleโ€™s rewrite engine has improved since August 2021, in terms of the truly bizarre edge cases. Hereโ€™s a funny one, though, from Google.com itself:

Even Google thinks that Google said โ€œGoogleโ€ too many times in this <title> tag. I suspect the rewrite engine flagged the word โ€œGoogleโ€ as redundant, but Iโ€™d definitely call this a misfire.

A more nuanced pie chart

I made myself a to-do of creating a โ€œpie chart with nuance,โ€ and I now realize thatโ€™s impossible. So, hereโ€™s a pie chart thatโ€™s slightly less misleading. Many rewrites are hard to categorize and count, but letโ€™s take a look at the data if we carve out the truncation scenarios (all three) and the additions:

Separating truncations and additions, weโ€™re left with about 30% of <title> tags being rewritten in our data set. Keep in mind that many of these rewrites are minor and some probably involve forms of truncation and/or addition that were difficult to detect programmatically.

Flipping this around, we have 70% of titles not being rewritten. How do we reconcile that with Googleโ€™s 87%? It could just be a function of the data set, but letโ€™s carefully re-read that quote from the beginning of the post:

Based on your feedback, we made changes to our system which means that title elements are now used around 87% of the time, rather than around 80% before.

Note the highlighted text โ€” Google is specifically saying that they used the <title> element/tag 87% of the time. They may have subtracted from, added to, or slightly modified that original data (they donโ€™t really say). So, the 13% of cases here is likely only when Google is pulling the display title in search from some other area of the page (body content, headers, etc.).

As to the bigger question of how much Google toned down rewrites after the initial outcry, itโ€™s difficult to measure precisely, but Iโ€™d say โ€œNot very much.โ€ It does appear that some edge cases โ€” including mishandling of parentheses and brackets โ€” did improve, and I think Google turned down the volume a bit overall, but changes to titles remain fairly common and the reasons for these changes are similar to August 2021.