More than half of the page titles will be rewritten by search engines, what is the meaning of writing titles carefully?

📅 2026-11-01 👁️ 1

It is said that more than 60% of page titles have been rewritten by search engines, does it still make sense to write titles and TDK carefully in this case - it does, but the position of the meaning has changed: the website cannot control what is ultimately displayed, but it controls how accurate the original material fed to the rewriting algorithm is.Search Engine Journal reported a study that statistics show Google changed More Than 61 Percent of title tags, the sample size was 30,000 URLs (report date January 31, 2022).This number indicates that rewriting is common, rather than 'useless if written'.

How is this number of 60 measured

The method of this research is consistent: grab a batch of page title tags filled in by themselves, then compare them with the actual titles displayed on the search results page, and count the proportion of discrepancies. The sample of 30,000 URL belongs to a single batch of observation values, and the rewriting rates under different site types and different query words are very different, and cannot be taken as a definite number for your own site.

Pay attention to the fact that it counts the 'discrepancy between the display title and the source title', among which many substitutions are equivalent: added site name, removed redundant words, and truncated long titles into more readable phrases. What truly needs to be paid attention to is the part that deviates from the original meaning after substitution.

What kind of title is more likely to be replaced

The proportion of three types of poor input quality is higher: the title is too long and is truncated into a sentence; the main keywords are buried at the end of the sentence, and the first half is the column name or modifier; the title does not match the content of the text, and the search side would rather generate a sentence from the anchor text or the text.

In reverse, the extent of replacement of a clearly written title is usually small, at most adding a brand suffix. When writing a title, put the key words in the first half sentence, control the length, and align with the content of the article, and these three things cost very little, the benefit is to make the rewriting algorithm have no reason to replace.

Which half does AI-generated TDK and manually proofread manage?

AnQiCMS's AI capabilities include AI auto-tagging and AI TDK: Automatically generate tags, images, and title keywords when creating articles, as well as AI writing assistance provided since April 2023. Their proper use is to draft initially - so that when content is batch-launched, the title, keywords, and descriptions are not left blank or written by the editor at will.

The human proofreading process manages three other things: what the content is supposed to answer, what to say externally, and whether it is consistent with the existing expression standards of the site. Leave a review position before going online, which is more time-saving than relying on rewriting to补救 after publication.

The display materials outside the title can also be handed over to the system for unified tone: the on-site anchor text management and keyword library are responsible for fixing the spelling of core words, avoiding four different ways of expressing the same concept in different columns. The link promotion and site map generation determine how quickly a new page can be read, and they, along with the title quality, affect the display results.

When is the rewritten result harmful?

Three situations are worth revising: when the brand name is replaced with a column name, the user cannot recognize the source; when the title is changed to a question or a combination of numbers, which does not match the actual content of the page; when the original compliant expression is truncated into a phrase containing absolute expressions, which brings advertising law risks. When these three types occur, rewrite the title and the first paragraph of the page to make it easier to extract, which is more effective than submitting feedback.

Most of the other replacements belong to the optimization of the display layer and do not need to adjust the title repeatedly.

Which pages are worth keeping the old title unchanged

  • Old articles with stable search sources for a long time: when the title changes, the comparison relationship of historical results is broken, confirm the consistency of the landing page before changing;
  • Brand page and official event page: There are compliance requirements for expression, it is better to be a little longer to maintain completeness;
  • For the pages with external references: The external anchor text is built around the old title, and the cost of replacement is high.

For this kind of page, the adjustment approach is to add new content instead of changing the old title, write the new question into a new page, and let the old title continue to accommodate the original query.

Common questions

Since it will be rewritten, do you still need to fill 60 characters?Do not fill with too many words. The part that exceeds the display window will be truncated by default, and it is more useful to write a clear theme than to fill length.

Will the title generated by AI be judged as low-quality content?The judgment depends on the match between the content and the query, not the generation method; the combination of automatic generation plus人工proofreading is more stable than the bulk generation of the entire site.

Should the brand name be included in the title?Yes. The search side often adds the site name automatically. It is more controllable to write a standardized「theme word + brand」structure yourself than to let it be replaced arbitrarily.

Related articles

How to judge the source of free circulation cracked themes that inject backdoors into installation packages

Crack themes and plugins by writing backdoors into the installation package, once the site is installed, it will be contaminated. In historical cases, the CryptoPHP backdoor in 2014 affected more than 23,000 sites. To judge the source, we need to look at three points: whether the authorization chain can be traced back to the author, whether there are non-source code obfuscated files in the installation package, and whether there are unknown outgoing requests after going online. Corporate websites should manage content defense lines for text but not for third-party code, and theme files still need external verification.

2026-11-01

The date when the content system stopped supporting was written into the announcement, and how to calculate the cost of upgrading or migrating the old station.

After the end of support, the system will no longer have official patches, and publicly disclosed vulnerabilities will accumulate more and more, the cost estimate for the longevity plan must be based on no one guaranteeing the bottom line. The two accounts of the old site should be calculated separately: for on-site upgrades, check module compatibility and template regression; for overall migration, consider content export, field mapping, and the search engine taking over old addresses. Regardless of which path is taken, make a complete backup including static files before taking action, and reserve a rollback practice window.

2026-11-01

Writing the background as code for website building, when the development is convenient, will the editor have difficulty taking over?

Payload This kind of code-first open-source system writes the content model and backend interface into the code, and the development team completes field modifications and view additions within the version control library. The cost is that daily adjustments on the editing side have to go through the release process. The difference between the two methods lies in the division of labor model rather than the presence or absence of features: editing templates with the assistance of visual templates, configuring fields with the content model, can move the modification path of non-technical roles from the code back to the backend.

2026-11-01

The content system that does not use a database, the savings in maintenance is real, but what is taken away?

The content storage system that stores content in files instead of database tables, eliminating a layer of database maintenance, and also makes versioning and migration simpler; the cost is in concurrent writing and large-scale retrieval. Statamic provides dual forms from flat files to database, selling the cost of switching rather than both being optimal. For enterprise websites with diverse content types and multiple writing parties, they will eventually return to a database form with backup and recovery capabilities.

2026-11-01

Which rule does the AI crawler read first when entering a site, and how do robots.txt and llms.txt divide labor?

When an AI crawler enters a site, it first reads the crawling permit, then the overview file: robots.txt determines which agents can crawl and which directories cannot be entered; llms.txt determines the order in which the model understands the site content after entering. Both are not mutually exclusive; one manages access permissions, and the other manages the reading path. They need to be configured separately within the site, and each should be verified once after changes.

2026-11-01

About 10% of the sites that use llms.txt, what problem does this file solve?

A third-party study covering nearly 300,000 domains shows that sites with llms.txt only account for about 10.13%, and it was not found that it can improve the AI citation effect. This does not negate its use: the overview file solves the problem of the reading path of the model after entering the site, allowing the model not to traverse the entire site itself, it does not replace the crawling license, nor does it guarantee that it will be cited.

2026-11-01

Is there any other way to send new links to search engines besides submitting a sitemap

The site map is a full list, an active notification is an update event, and the two solve different time differences. The role of IndexNow and similar protocols is to inform search engines as soon as the content is created or modified, with link push management handling this within the site and connecting to Baidu push and Bing channels separately; the site map is still responsible for covering historical pages and column structure.

2026-11-01

The price difference for the beginner package of hosting website starts from a few dollars a month to twenty or more dollars, what do you get for the price difference?

The entry-level pricing for mainstream hosting and website building platforms is roughly between $4 to $25 per month, the price difference mainly involves three aspects: whether the platform ads are removed, whether e-commerce and form components are unlocked, and what the collaboration seats and traffic limits are. Self-hosted content management systems have a different billing structure, where the host and runtime environment are in your own hands, and the functionality is provided by the system itself, with memory usage and deployment methods directly affecting resource costs.

2026-11-01