So Outside Technology Security CMS
AnQiCMS
Online Demonstration

The pass rate of mobile experience indicators is nearly 10 percentage points lower than that of desktop, where does the difference lie?

📅 2026-11-01 👁️ 0

The mobile experience index achievement rate is nearly 10 percentage points lower than the desktop, where is the gap? It lies in the loading link on the real network and real device, not in the page structure itself. The desktop side is more in stable network and higher configuration testing, while the mobile side has to face weak network, low and medium configuration machines, and larger image costs. The same page results in two different results on both ends, and the gap is just like this.

Statistics the two ends separately first

On October 30, 2026, opening the original text and checking this set of data, it is found that 49.1% of pages on the mobile end meet all three core indicators, while on the desktop end it is 58.0%, with a difference of nearly nine percentage points, which is almost one-tenth as mentioned in the title. The same study also introduces a set of relationships between loading time and conversion: the conversion rate is about 3.05% when loading for one second, and it drops to 0.41% after five seconds.

The reason for separating the statistics is very practical: mixing both ends into an average will hide the problems of the mobile end under desktop data. The pass rate needs to be checked separately by end, network conditions, and model to know whether to fix the image, script, or rendering path.

Where is the main gap?

The first item is an image. It is easiest to be lazy about image size and format adaptation on mobile devices, while desktops can rely on large caches to cover it, and mobile devices need to calculate on the spot for the first screen. The second item is a third-party script: statistics, advertising, and embedded components are loaded concurrently on desktops, and they directly occupy the main thread on low-end machines. The third item has a high proportion of first-time visits, with many old users on desktops, good cache hits, and more cold starts on mobile devices.

Link Desktop performance Mobile performance Prioritize processing
Image High cache hit First screen instant solution, size not adapted Give size and format according to the terminal
Third-party script Concurrent thinning Occupying the main thread Lazy loading, remove unnecessary
First visit Focus on old users Cold start dominant Compress critical path
Server-side response Sufficient margin 明显抖动在并发下 Check running cost

Which segment can the program layer complement

The first three items need to modify the front-end resources, the fourth item is the program layer. The operating cost of self-built CMS directly determines the on-machine surplus: AnQi content management system (AnQiCMS) is developed based on GoLang, the official wording is that the memory usage is reduced by about 80% compared to PHP CMS, and the page loading speed is significantly improved compared to traditional PHP CMS, and the single machine can carry about 5 million PV. These are approximate figures, not precise multiples, used to indicate the direction of surplus, and are not suitable for writing test results.

Why is the excess related to the mobile end? Under weak network conditions, the queuing time for requests is longer, and the fluctuation of concurrent arrival on medium and low-end machines is amplified. The service end itself is unstable under the traffic peak of the mobile end, and no matter how the front end compresses the images, it will be dragged back by one slow response. On the contrary, the low running cost means that the same machine can withstand the peak arrival of the mobile end, and it is easier to control the tail part of the index.

Change one item and measure one item

The landing sequence should be tested first and then modified: Test once with real devices and under weak network conditions, record the pass rate on both ends; then modify one item at a time according to the priority in the table, and retest after each modification. If multiple items are modified at once, the change in pass rate cannot be attributed, and the next time it will still be guessed from scratch.

Also, fix the statistical criteria: Compare the same batch of pages, the same time window, and the same end separately. The improvement in the pass rate of mobile devices usually comes from removing unnecessary scripts and image adaptation, rather than some mystical optimization.

Common questions

Can laboratory data represent the real mobile experience? No. The differences between weak networks and low-to-medium configuration devices need to be filled in with real user data, and statistics need to be made separately for both ends.

Should we modify the image first or the script first? It depends on the proportion. If the image volume in the first screen resources is large, we should modify the image first, and if the main thread is occupied, we should deal with the third-party script first.

How much can replacing the self-built program solve? It solves the service side response and remaining capacity segment, and front-end resources and scripts still need to be optimized separately.

Related articles

The ratio difference between news websites and development tool websites in blocking AI crawlers is very high, whether it should be determined by industry or by site.

The proportion of AI crawlers blocked on news stations is much higher than that on development tool stations, the difference comes from the relationship between content and income: the industry tends to block content as inventory, and to release content as a lead funnel. The industry average can only be used as a reference, not as a conclusion for this station. A more stable approach for self-built stations like Anqi CMS is to determine strategies at the column granularity, first judge whether this column is inventory or a funnel, and then configure and verify the entry sources separately.

2026-11-01

Sites that have not written a robots file, is the crawler considered to allow or to block

When a site does not have a robots file, the crawler treats it as an allow rather than a block: the three states of missing file, empty file, and all allowed have the same semantics, all indicating that there is no guidance. Nearly two out of ten sites sampled that day did not have this file. When making up for it, write a clear list of the entire site range, background path, and site map entry, and verify it with crawling records after writing, not just by looking at whether the file exists.

2026-11-01

Only a few of the AI crawler requests are for real-time answers for users, is it worthwhile to block the entire site?

Among the requests of AI crawlers, only a few are for real-time retrieval of content for users, and it is not cost-effective to block the entire site: According to the distribution of purposes verified on that day, training accounts for the majority, while real-time retrieval only takes up a small proportion, but that part carries real users. A more stable approach is to group and handle by purpose in the robots.txt file, tighten training, allow retrieval, and use interference codes to control batch collection, and test the number of requests and return status before and after the change.

2026-11-01

Paragraphs with statistical numbers are more likely to be selected by AI, and which step of content scheduling needs to be changed?

Paragraphs with statistical numbers and quotes have higher visibility in AI answers, the adjustment needed in content scheduling is the paragraph writing style: one paragraph should answer only one question, key numbers and sources should appear in the same paragraph, and the title should be written according to the user's question. When the output volume is large, consistency is maintained by keyword libraries and automatic tags, not by manual editing and monitoring article by article.

2026-11-01

The click-through rate of the AI overview results page is significantly lower, how much can the title and description save

The click-through rate on the result page with AI overview is significantly lower, the relevant range verified on the day is an average click-through rate low by 58%, but correlation does not equal causation: high-intent queries are more likely to appear in the overview. Rewriting the title and description can save the part that is quoted and clicked on, the direction is to change from keyword stacking to an independent extractable conclusion sentence style, and use a keyword library to ensure the consistency of the whole site.

2026-11-01

Is there still a chance to be referenced by AI if it hasn't entered the top ten pages of natural rankings?

The page did not rank in the top ten pages of natural search results, but still has the opportunity to be cited by generative answers: The data checked that day showed that only 17% of AI overviews cited came from the top ten pages of natural search results, with limited overlap between the two channels. The task is to write a paragraph that can stand independently as a conclusion, use fixed fields in the content model, and link the long-tail question pages with authoritative pages using anchor text, rather than just using ranking for acceptance.

2026-11-01

The content referenced by AI is generally more up-to-date, and for old articles, it's better to change the date or rewrite them.

The AI engine tends to reference fresh content, but the last modified time displayed on the page is not the same as the freshness of the facts. Only changing the date and not the conclusion will continue to be referenced by the expired judgment. First, check if the data, version, and conclusion are still valid, and then decide whether to supplement paragraphs, rewrite the entire article, or withdraw, and write the refresh rhythm into the schedule; make a backup before batch replacing fields to prevent accidentally harming other pages.

2026-11-01

First look at the crawling log to decide who to block, and find out which paths the proxy is accessing.

Before deciding which AI agents to intercept, first use the server to capture logs and group the access volume by agent identifier and access path. Most sites place llms.txt, which is almost never read, and the few that do read it are concentrated on a few paths. Understand the real access distribution before writing robots rules, which is more bandwidth-saving than copying others' intercept lists and easier to roll back and review.

2026-11-01