It wouldn't be outrageous to assert that page titles are the most important information that most searchers take into consideration when skimming SERPs. Their significance is underscored by the fact that on all major search engines, these titles aren't just descriptions but are actually the links that users click to go to a website.
With over a decade of SEO practices now under our collective Web belt, website owners and Internet marketers have come to understand a few key generalities about what makes a good headline and helps a site stand out amongst its competitors, and thus, what drives higher click-through rates (CTR). For example, we know that titles need to be descriptive of the page content; they need to be unique and not repeated elsewhere on a site; they need to avoid keyword stuffing, and it doesn't hurt to add some branding, either.
But just because we know about some of these best practices for page-title generation, that doesn't mean that everyone follows them -- or that they work in every situation.
Google recently came out and dropped a not-entirely-shocking bombshell: Sometimes its algorithms may change the page title that webmasters designate for a site.
While Google asserts that it has always advised people "to write unique, descriptive page titles," they're apparently treated like meta descriptions, that is to say, more as suggestions than anything else. This is because the company has found that some titles generated by webmasters may not be the best options (the "most optimized," if you will), and in these cases the algorithm will "generate alternative titles to make it easier for our users to recognize relevant pages."
Despite the fact that Google primarily looks at the <title> tags specified in a site's HTML markup when deciding on the best title, the given title is not always used, and, theoretically, the webmaster/site owner may never even know; although, Google does say that it tries to notify webmasters when it discovers "titles that can be improved on their websites."
To be fair, Google has the best of intentions at heart, and the alternative titles that are selected are done so based on testing to determine the title most relevant to the query. In the end, this "can substantially improve the clickthrough rate to the result," according to Google.
However, relevancy is only the reason for alternative titles "about half of the time." The other half is for pages that (A) don't have titles, (B) specify non-descriptive titles (such as simply, "Home"), (C) use the same title (or just minor variations) on most or all of a website's pages or (D) are unnecessarily long or hard to read. In these instances, Google is essentially cleaning up poorly concocted titles and replacing them with algorithmically approved alternatives that are more informative and helpful to searchers based on their queries.
Unfortunately, as with most information about Google's algorithm, there isn't much available when it comes to how the company determines the best alternative titles for a site. Typically these new Google-created titles originate from words pulled out of the content on the page, which is much the same tactic the algorithm uses when it crafts its own page descriptions for SERPs.
Though this may come across as somewhat convoluted, the goal on Google's part is simple: to help users by providing them with the most relevant information about the content of a Web page and, in turn, increase CTRs for the sites listed on SERPs. By helping to optimize your titles for better results, Google is creating a win-win situation for site owners/webmasters and searchers alike.
And it's important to remember that Google reserves the right to change titles as it sees fit, so if you don't like the idea of the company toying with the information you present, your only option is to optimize page titles yourself. The good news is that Google tells you how to best do that in its Help Center.