Articles written by GPT AI for scenarios have overly obvious machine-like characteristics, and it is truly difficult to build a set of highly readable copy. This is due to GPT’s own characteristics: GPT solves the problem of answering, not writing articles, so its writing methods and writing style are fixed. If you want it to write good-looking, easy-to-read articles, you need to set many conditions, so we do not really recommend using GPT as the first choice.
[baike]AI dedicated to copywriting[/baike]Huoshan Writing is an AI specifically for copywriting. Huoshan Writing is under ByteDance and is trained on copy from articles and videos on Douyin, Toutiao, Xigua, etc., learning the writing methods of excellent content with high likes and high interaction. Its goal is to write high-quality copy and improve readability and interactivity, so using Huoshan Writing to write articles is the right approach.
From its interface, Huoshan AI clearly has strong copywriting-tool characteristics: all kinds of settings are there to optimize the copy, and the process is the same. For example, when you enter a title, as soon as you type a few characters it starts helping you construct the following keywords and tells you how to write in a more compelling way—for example, how to design the first 5 seconds according to the golden rule. After you enter a topic, it will help you generate copy; unless otherwise specified, it will generate different copy structures according to the characteristics of the title.
The copy generated by Huoshan AI is visualizable: it tells you the source of the content, which published content the current copy is based on, and what the data for that content looks like.
[baike]How search engines see it[/baike]Baidu does not support using AI-written articles to update websites. For this reason, the moderator once consulted the product customer service of ERNIE Bot and Baidu Search. The ERNIE Bot customer service meant that it should be best not to use it for updates, because the data itself mainly comes from search engines, and the generated articles are very easy to identify. The response from Baidu Search was even simpler: because it was an online conversation, their reply was that they do not recommend using AI for updates.
As for Google, before launching Gemini they said they would not accept updates with AI-type content, but after Gemini was released they said that good AI-type articles are worth recommending.
Algorithm loopholes
When analyzing this issue, the moderator discovered a little trick: search engines seem to treat scarce content specially. For example, the moderator found a website, douhao.com, whose content is very easy to tell is AI-generated—the formatting is obvious, and the content cannot solve practical problems, so user behavior data is definitely not good. But its advantage lies in the irreplaceability of the content: there is no similar content competing with it. Baidu’s characteristic is not “better to have nothing than something inferior”; as long as your content is scarce, it will be displayed to users first. Only when there is a lot of similar content will it make comparisons. This is also a small loophole.
[baike]Content timeliness[/baike]Huoshan Writing does not belong to the search engine camp; its writing is very natural and highly readable, so even search engines cannot tell whether it was generated by AI. If it is used to update a website, the moderator’s top recommendation is Huoshan.
In addition, search engines have a strong lag, especially Baidu. For the latest content, Baidu needs a long process of analysis, indexing, and display, so it is very hard to find content about newly launched products. For example, when Sora first came out, many people may have written articles, but Baidu had not yet displayed them, so you could not find them. Huoshan is different: it ingests high-quality content in real time—the front end is an algorithm, and the back end is a real-time expanding database. The difference is immediately obvious.
[baike]How to write good copy[/baike]Writing good copy with Huoshan Writing is also not one-click generation; it likewise needs optimization, but this process is very simple and visual. After generating a basic copy, you can set specific requirements through full-text optimization, including writing purpose, target audience, tone and attitude, etc.
[baike]Experts use custom GPTs[/baike]Custom GPTs are actually what the moderator recommends more. The knowledge base is your own, and the writing style and writing techniques can be set freely, but the premise is that you need to understand article structure very well and understand copywriting techniques very well, and then set the content source as the knowledge base. In this way, you can generate content with a unique style. But for novice users it is relatively complicated, and currently only ChatGPT supports it relatively well.
The GPTs I have trained will not use structures like “first, then, finally, conclusion”; moreover, they will use many writing techniques, cases, data, effects, and stories, so readability is very strong. As shown, I added knowledge base files when setting it up, and in the text settings I required it to learn and use the style I commonly use.
At first, you will use very simple AI tools to create content—that is how I came through it. For example, I first used ERNIE Bot to quickly generate a large number of articles to update websites, but the website almost died. Later I used Huoshan and it was much better, but it could not set my independent style. In the end, it was GPTs that could meet my needs. This is a pattern and also a process of learning and evolution. |