So Outside Technology Security CMS
AnQiCMS
Online Demonstration

Only about forty percent of the pages cited by AI overview are in the top ten, should the external links be adjusted?

📅 2026-11-01 👁️ 0

The proportion of pages referenced by AI overview and ranked in the top ten of search results has dropped significantly over the past year, should the external link strategy be adjusted accordingly? Don't move the external link plan yet. This study covering more than 860,000 search result pages shows that the proportion of pages referenced by AI overview and ranked in the top ten has dropped to 38%, compared to 76% a year ago. The decline indicates that 'good ranking' and 'being referenced' are beginning to decouple, but it does not mean that existing ranking work has nothing to do with references.

What do the changes in numbers indicate?

There are two interpretations for the decrease in overlap, and the data itself cannot distinguish between them. One is the expansion of the reference source pool: the model begins to reference more pages that are not in the top but can be abstracted in content, and naturally these pages are not in the top ten. The other is the change in ranking results: the difference between the standard of selecting the top results and the overview reference has become larger, and ranking high in the same query and being cited have become two independent things.

Regardless of any explanation, it can be confirmed that the same thing is: taking 'improving ranking' as a synonym for 'improving the probability of being cited' is no longer valid this year. Abolishing the external link plan is equivalent to making a significant decision based on a curve that has not yet determined the cause.

Why would the overlap decrease?

The overview references are more about whether the content can be directly extracted. For the same query, the pages ranked at the top are often the homepage, category page, or comprehensive long article, which bear the function of entry; while the model is more likely to extract paragraphs that are self-contained, have clear conclusions, and are accompanied by specific data and judgment conditions. The forms of these two types of pages are inherently different.

Additionally, the sorting of search results is greatly influenced by authority and link distribution, while citation extraction is more closely related to the answerability at the text level. This is why many highly cited pages do not make it into the top ten: they write more directly on specific issues, but are not necessarily the sites that rank highest in the overall sorting of the query.

Which external link plans do not need to be changed

Three things do not need to be changed. First, external links pointing to reliable sources and industry media are still valid, affecting the credibility and exposure of the site, and should not be cut off due to changes in the scope of reference. Second, consistency between content updates and natural link acquisition should be maintained, as sudden changes in the rhythm of external links are most likely to leave a cost in rankings. Third, technically accessible: maintain the pseudo-static URL rules, 301 redirects, site maps, and link submissions to search engines as usual, as pages that cannot enter the index pool have no meaning to be cited.

The assessment method needs to be changed: external link work continues to be evaluated based on ranking and traffic, the citation situation is observed separately with a separate line, and the two sets of indicators should not be mixed into one conclusion.

Two things can be done first within the site.

The first thing is to tighten the scope of keywords. The AnQiCMS keyword library is used for centralized management of a set of SEO vocabulary, and anchor text management allows keywords within the site to automatically form links. Both use the same vocabulary list, which can make the internal links under the same theme clear: when the model reads a page, it can judge from the anchor text where this word has a more complete explanation within the site, and readers can also follow the chain to reach deeper pages.

The second thing is to change the main text into a form that can be quoted. Each subtitle should be written as a question, and the paragraph should contain the conclusion and conditions independently, with numbers and judgment criteria written at the beginning. These two adjustments do not affect external links, but they directly change the quality of the page being extracted and make it easier for citations and rankings to appear simultaneously.

Review the composition of sources quarterly to see where your site is referenced, which is more instructive for the next step than watching the overlap percentage weekly.

Common questions

Does the decrease in overlap mean that SEO is no longer useful?Not. The decline indicates that citations and rankings are no longer rising and falling together, ranking work still determines the entry traffic, and citation work relies more on the text form.

Should we make a batch of pages specifically for citations?We can first change the paragraph structure on existing pages and observe for a quarter. The new page dilutes the internal links of the existing theme, and the effect is often not as good as writing the conclusion clearly.

Should the anchor text be added more?Control by word table density. The same keyword should point to the same target repeatedly within the site, and stacking anchor text will interfere with reading and will not increase the probability of being extracted.

Related articles

The adoption rate statistics of two llms.txt files differ by more than a factor of two, so which one should be used to allocate work?

The adoption rate of llms.txt is the same, but the readings from different studies range from ten percent to nearly fifty percent, the gap mainly comes from the denominator rather than the conclusion. This article decomposes several statistics according to the sample size and what they can answer, and gives three indicators that are more important than the adoption rate: whether it has been read, whether the content is consistent with the site map, and whether the maintenance cost can be automated.

2026-11-01

The sites that block training crawlers are almost no longer cited, can this data be implemented directly?

A study covering thousands of domain names shows that almost no references can be obtained by blocking the domains captured by training class. This article explains why this set of readings cannot be directly applied: sample scope, industry composition, and category mixing will all contaminate the results, and a minimum experimental design for small-scale comparative observations on your own site is given.

2026-11-01

Pages with ads are blocked by default for training crawler, what configuration line should the new domain check?

The newly connected domain blocks training and proxy crawlers by default on the page displaying ads, while search crawlers are still allowed. The verification should be done in three layers: which domains and page types it affects, the user agent names written in the configuration, and the actual visits and rejection results in the access logs. This article also explains the division of labor between access control, site map, and llms.txt.

2026-11-01

The big version at the end of the year comes with a new default theme, and before the upgrade, the test site should be scheduled according to which nodes.

The big version at the end of the year is released with the new default theme, the testing station schedule needs to be reversed according to the official nodes: in the test version phase, only compatibility detection is done, in the candidate version phase, the regression of four types of surfaces is completed, and on the day of the formal version, it is grayed out first and then fully released. This article gives the five-step order on the day of the upgrade, three options for the rollback path, and the check items for the pseudo-static rules and backup.

2026-11-01

Two weeks after the update, the rankings are still fluctuating. Should we change the content in this round?

Completion of push does not equal stable ranking. This round of garbage update took two weeks to complete, and the official also indicated that recovery may take a month. This article gives the order of handling within two weeks after completion: first, fix the baseline and the timeline of self-change, and then judge whether to change the content according to three types of signals. How to minimize the scope of changes when batch adjustments are needed.

2026-11-01

Exporting the entire old station into a package and then importing it, will the attachments fall outside?

The packaged file exported from the entire station usually contains media files, whether the attachments will fall outside when imported across systems depends on whether the path and fields can correspond. This article compares the three import routes of packaging, table, and compressed package, respectively, what they take away, explains where the upload volume limit is stuck at each level, and gives the three acceptance and rollback preparations after migration.

2026-11-01

Can a lightweight memory PHP program withstand the corporate station, what indicators to look at

Can a lightweight memory PHP website program withstand an enterprise station? It needs to be evaluated based on three indicators: the number of data tables and the size of the code package only reflect the installation surface, the resident memory water level, single machine concurrency carrying capacity, and static structure determine the peak performance. This article provides a set of verification order, first read the volume, then run pressure test, and finally draw conclusions according to your own site configuration, and attach the memory aperture of the Go technology stack for reference.

2026-11-01

Who manages the server for self-built background and hosting platform, and where is the difference in maintenance working hours?

The difference in server division between self-built CMS backend and hosting platform is concentrated in the three steps of patching, backup acceptance, and upgrade rollback. This article compares each item with the patch frequency and labor cost range given by the operation and maintenance cost study, explaining that hosting does not mean zero labor and self-building does not mean all manual, and provides a method for calculating maintenance hours into annual costs.

2026-11-01