While ChatGPT is a revolutionary technology that can produce content, there are several points to keep in mind to produce good SEO content. In this post, I share tips for SEO with ChatGPT based on my experiences writing blog posts.
What is ChatGPT?
ChatGPT version 3.5 attracted 1 million users in one week becoming the greatest technical innovation to spread so fast. ChatGPT version 4, also called ChatGPT Plus was released as a subscription offering with a monthly fee of US$20. The advantages of ChatGPT4 are that it has an even bigger model, can access internet links and provide better answers. The system guarantees access with certain constraints. Developers also have created several plugins that enhance ChatGPT’s functionality.
Large Language Model (LLM) for ChatGPT 3.5 utilizes vast amounts of data from Wikipedia, books, and social media. This enables the platform to produce a chat session that closely resembles human language and is culturally appropriate. While ChatGPT can provide coherent relevant responses, it lacks deep understanding, as it is trying to generate the next word based on the previous words. In addition, the responses may not be consistent and may not be correct sometimes.
WebChatGPT plugin
WebChatGPT, a plugin for ChatGPT, enables it to process the latest information on the Internet. It adds relevant web search results based on the prompt as the search phrase, enabling the platform to give responses based on the links. If the links are not relevant, ChatGPT will provide a response based on its training data.
Improving the ChatGPT response
We need to focus on the following tips for SEO with ChatGPT.
- Fact-checking the output is crucial to ensure the accuracy and trustworthiness of the content.
- Copy editing of the output is necessary to remove duplicates and to fix grammatical or structural errors. Ensure that the content flows well and is coherent. Tools like Grammarly can be helpful.
- SEO check using AIOSEO or Yoast SEO helps in analyzing keyword usage, search engine ranking, readability analysis…etc.
Prompt Engineering Example
Preparing a series of prompts to achieve a desired outcome is called Prompt Engineering. Let me explain with an example of generating a blog post for the review of Robinhood, a user-friendly platform for millennials investing in the USA.
First Prompt
Take a look at the prompt text along with the screenshot below.
“Write a 1000 word SEO blog post on Robinhood Review: The post is meant for young investors. You can use headings, bullet lists to make it easy to skim for mobile readers. Include references to third party reviews such as Android Apps reviews. The keyword to be optimized is: Robinhood review”
The following figure shows the Initial part of the response.
Second Prompt
The response to the first prompt was 361 words. Since I didn’t meet the word count, I can add another prompt to expand a paragraph of interest. Below is the prompt for expanding the paragraph titled ‘What is Robinhood?’
“Expand on this paragraph about Robinhood review: Robinhood is a mobile-first trading platform that offers free stock, options, ETF, and cryptocurrency trades. With a primary focus on democratizing finance for all, it has become a favourite among millennials and Gen Z investors.”
After updating the response from the second prompt, the word count increased to 561. This process needs to be repeated to reach a word count of 1000 words.
You can see the annotated log of the above interaction with ChatGPT for this example at this link.
My Learnings with ChatGPT
My learnings, based on my experience of writing 7 blog posts are given below.
- ChatGPT can generate the copy for the blog in 2 to 5 minutes, while manually researching the content and writing the blog can take 1 to 2 days.
- While the platform produces content meeting the typical expectations for a blog post, SEO practices compliance such as keyword density, and readability improvements need to be done manually.
- When using WebChatGPT, sometimes the search results obtained may not be relevant to the topic. In that case, use a different search engine such as Bing.
- The blog post might use numerical information, such as dates and revenues, pulled from search links. These need checking using authoritative primary sources.
- When we expand the answer to the first prompt for some paragraphs, potential duplications may occur which need copy editing.
- The answer may contain words that indicate uncertainty because of the nature of ChatGPT. These need fixing by consulting the authoritative primary sources.
- I also recommend running the final output through an AI detection tool like GPTzero to ensure that the AI content is less than 33%.
To sum up, while ChatGPT gives a big productivity boost for content writers, the copy needs to be carefully reviewed and facts need to be verified for a proper SEO blog post.
Did you try ChatGPT? Do you have a question or comment? Let me know.
