My technical blog day PV dropped from 2000 to 1000: AI is eating search traffic

Foreword: This is not a case, it is the collective dilemma of all back-end technology bloggers

My blog focuses on back-end operation and maintenance content such as PHP, GO, MySQL, Redis, Linux, Nginx, system architecture, and network optimization, and has been in operation for more than 10 years. In the past year, I really realized that the entire technology industry has quietly changed.

In the first two years, the daily PV of the blog is stable at 1500-2000, and write a GIN row pit, mysql optimization, nginx The just-needed articles configured can easily break 2000 in a single day; but now the normalized low peak, the average daily PV is only 1,000, and the traffic is directly reduced by three-quarters.

At first I have been self-doubting: Is the content I wrote too old? Is the knowledge out of date? Or am I too few updates?

After reviewing Google Analytics traffic data, I figured it out completely:It’s not that my article is not easy to use, it is the way the programmer finds the answer, which has been completely changed by the AI.

The whole article is all the real GA data of my background, which does not create anxiety or blow AI myths. Combined with my true feelings of using TRAE and Tongyi Spirit code every day, I will talk about the common difficulties of all back-end developers and technology bloggers.

1. GA real shot data: visually see the flow of the cliff-like decline

This article only sees the most realIn the past 30 days, the whole station traffic, the data source is Google Analytics.

Real shot status:

1. Normal in previous years: the average daily PV1500-2000 on working days, most readers came in from Baidu and Google search keywords;

2. Now the normal state of the past 30 days: the average daily stability is only 1000PV, there is no stop and no holiday off-season, which is a real long-term decline.

In the past, the programmer encountered an error, and the process was very fixed: copy the error message → search engine search → click to open a blog to find a solution.

Now the process of most people is directly simplified: copy errors → open trae/tongyi spirit code → AI directly give runable code, troubleshooting scheme,If you don’t open a search engine in the whole process, you will naturally not open any technical blogs..

2. Core real hammer: the natural traffic of search engines continues to decrease

Many webmasters think that the traffic of the entire network website is falling, and everyone is the same.

I specially opened the GA traffic channel page to check, and I came to a very straightforward conclusion:

Only the search engine’s natural traffic is falling, and the rest of the channels are almost unchanged.In my search engine traffic, Bing actually surpassed Baidu. I feel that Bing is about to surpass Baidu in China? Haha! !

Combined with the current situation in the 2026 circle, it is completely reasonable:

1. Now programmers find code and arrange bugs, and the frequency of using AI tools has exceeded Baidu and Google;

2. After Google has its own AI answer, there are many technical keywords without clicking on the website, the page is directly out of the answer, and the blog is zero-click;

3. Small and medium-sized technology stations, the average search traffic has dropped by more than 60% in the past two years, which is completely in line with my blog data.

Subdivided, my blog traffic has fallen the worst, all of which are the basic content:

• PHP basic syntax, Laravel general usage tutorial
• Introduction to Go/Gin Basic Routing and Middleware
• Mysql regular sql writing, basic error solution
• Redis basic command, Linux/NGINX basic configuration

This kind of basic tutorial with standard answers and highly repetitive, search traffic is basically cut in half.

On the contrary, the traffic of high-level content is very stable: online difficult bug review, architecture selection, and server in-depth tuning.

3. I am also an AI user: I understand high efficiency and anxiety

I am not excluding AI programming, on the contrary, I am writing code every day, which is inseparable from TRAE and Tongyi spirit codes, and my daily workbench screenshots are attached, which are all real and self-use.

Objectively speaking, the convenience brought by AI, neither blow nor black:

✅Write CRUD interface, standardized configuration, basic script, the AI speed is far exceeding handwriting;

✅ General code troubleshooting, syntax error correction, environment configuration, AI is given a few seconds;

✅ Now there are Agent and Skill capabilities, project architecture, unified code specification, writing unit tests, simple code review, and even taking over the code of bad second-hand projects written by others, and AI can handle most of them.

This is also a complete explanation, why no one searched the basic tutorial:

In the past, I used to write a new interface, and I had to search for half a day of blog, copy code, and repeatedly debug it; now open TRAE to clarify the needs, and directly get the code that can run.

Because of this, I have been very internally used recently, and this is also my most real inner thought at the moment. I believe that many back-end programmers who have been deeply involved in many years have the same entanglement:

1. Novices rely on AI to quickly write code, and in just a few months, it seems that they can catch up with the framework and grammar experience I have accumulated for several years;
2. When I calm down the handwritten code, review the GIN source code, and sort out the underlying principles of GO, my heart is very calm and calm.
3. But if you don’t learn the new AI abilities such as Agent and Skill, you are afraid that you will not be able to keep up with the rhythm of the industry and gradually fall out of date;
4. After studying the AI tool with a single mind, I became impetuous again. I slowly lost my handwriting skills and thoroughly understand the bottom-level patience, and I couldn’t find the original state of writing code.

4. Transparent conclusion: The drop in traffic has brought a reminder to all programmers

1. Those who will be eliminated by AI are never developers who understand the bottom.

What AI can replace is all memory work: back API, memorization syntax, writing template code, and solving ordinary errors.

AI can never replace the ability to judge: it can be seen that the AI code is hidden and concurrent bugs, veto unreasonable architecture, optimize bad SQL, check online memory leaks, and make technical trade-offs in combination with business.

This is also the reason why I still insist on gnawing at the bottom and reviewing the source code of the framework:Learning the basics now is not for pure handwriting, but for the pits written by AI and the bottom of AI.

2. Technical blogs must be completely transformed

By the way, let’s talk about the follow-up writing mentality: I have never simply chased search traffic when I write a blog. Regardless of the flow of traffic, the core purpose of writing is always self-summary, technical refinement, and precipitation. The process of exporting and sorting knowledge points is essentially to force oneself to thoroughly understand technology, make up for shortcomings, and help self-growth.

It’s just that the current industry environment has changed, and the search traffic of basic tutorial articles has fallen sharply. I will not deliberately stop changing such content, but I will not cater to search again. The engine will write basic troubleshooting and command tutorials; follow-up writing still follows its original heart, takes into account self-precipitation, and synchronizes the creative content of the current AI industry.

In the follow-up, I will deeply cultivate the three types of content, and I will still focus on self-precipitation, and the flow is random:

• AI step on the pit record: TRAE / Tongyi spirit code to write code, hidden performance, safety, business bugs, review and avoid pits
• Comparison of man-machine code: handwriting bottom-level VS AI generates code, measured pros and cons, and consolidates one’s own skills
• High-level actual combat: architecture selection, online troubleshooting, Agent+Skill landing experience

3. Personal learning ratio: adapt to the industry, keep the heart

The core of this article only analyzes the cause of the decline in traffic, and simply talks about personal learning attitudes by the way: in order to follow the trend of AI, I will not lose the bottom-level skills of the back-end; I will not stubbornly write and refuse to improve AI.

Set the long-term learning ratio for yourself, and adapt to the current AI environment:

60% deep-floor foundation + 30% polishing AI human-machine synergy ability + 10% actual combat step on the pit review

5. Written at the end (return to the core of this article: the essence of the decline in traffic)

Combined with the two real shot data of GA’s total site traffic and search engine channel traffic in the past 30 days of this GA, the core conclusion of the whole text is summed up, which is also the only purpose of this article’s writing:

In the past, programmers competed: who wrote the code faster, who ordered more, and who knew the knowledge point well;

1. The drop in the traffic of this site is not caused by the decline in the quality of the content and the slack in the update;

2. Core roots: Developer Q&A channel transfer, from ‘Search Engine Search Blog’, to ‘AI Programming Tool Direct Answering Questions’;

3. Concentrated categories: php/go/mysql/linux/nginx standardization basic tutorial, general error report, configuration tutorial;

4. Traffic differentiation is obvious: the flow of superficial standardization technology has plummeted, and the traffic of high-level experience, underlying principles, and difficult troubleshooting content is stable;

5. Personal mentality: look down on the rise and fall of website traffic, and write blogs only to improve self-precipitation, not to cater to search engines to produce content.

AI just eats up the basic search traffic, just changes the communication channel of technical content, and will not change the value of deep-floor and precipitation technology.

Now the programmer competition: who can control the AI, identify the AI, and correct the AI.

AI just eats up the basic search traffic, and will never be able to eliminate developers who understand the bottom-level, architecture, and bottom-line problems.