Most marketing teams know how to create good title tags, but few actually find the time that is required to get the job done, especially at scale on big websites.

This AI-powered process for generating title tags takes into account the search intent of each page’s main target keyword and the purpose of the page, making it possible to generate high-quality title tags at scale.

Summary of the Process

The AI-driven process for generating title tags that I share in this guide consists of the following steps:

• Step 0 (Input Data): URL and keywords
• Step 1: Get additional keyword from current title tag
• Step 2: Get search intents for keywords
• Step 3: Get purpose of page
• Step 4: Select the best keyword
• Step 5: Create title tag

Make it repeatable and scalable

I use a tool called PromptMate for all of my AI processes, as it allows me to connect to various LLMs and AI systems and to import data from different sources.

With PromptMate, we can make this process scalable and repeatable, meaning that we can create title tags for thousands of URLs at once, or create them regularly for new pages or pages that require optimisation, always using the same process.

In some of the the prompts in this process, you will find placeholders, which are marked by double square brackets, and they either reference the output of previous prompts, or data from external sources that PromptMate can automatically fetch.

This guide is not supposed to be about the tools used, but rather about the process, so you can use it with any other software that is similar to PromptMate, or you can build your own solution.

Disclaimer: I am a co-founder of the company that develops PromptMate.

Step 0 (Input Data): URL and keywords

This process is designed for pages that are already live and thus have a URL that is publicly reachable. If you want to use the process for content that has not been published yet, you can easily adjust it accordingly.

The process also integrates keyword data – The better your keyword input is, the better your results are likely to be. But, even if you have no keyword data for your URLs, this process will still work for you, so don’t despair.

If you decide to import keyword data, it could be from any of the following sources, among others:

• Manual keyword research
• AI-supported keyword research
• Google Search Console
• Rank tracking tools
• etc.

I normally use GSC data if I don’t have the results of a decent keyword research at hand.

Step 1: Get additional keyword from current title tag

This step serves as a fallback option in case we don’t have any keywords that we can use in our input data or in case we have doubts about the quality of our imported keyword data, e.g. when we use GSC as a data source.

The prompt for this step is:

Which popular search keyword does a page with the following title tag provide the best answer for?
Title tag: [[Website content::title]]
As output, please provide only the keyword.

This prompt takes the current title tag of the page as input and asks the AI system to identify a popular search keyword that the page provides the best answer for based on the title tag. The output is a single keyword without any additional information.

Step 2: Get search intents for keywords

This step of the process analyses Google search results for all potential focus keywords of the page and deduces the search intent of each keyword from the search result page content. If you have fewer than three keywords per page, you can use the value “null” for the fields that you cannot fill. Alternatively, you can adjust the prompt, so that it uses more or fewer keyword variables, depending on how many keyword options you want to provide per URL.

The prompt for this step is:

In the following, I will give you a list of four keywords and their Google search results. For each keyword, please deduce the users' search intent from the corresponding Google search result. For your analysis, please ignore the keyword itself and consider only the Google search result.
If a keyword is "null", please do not analyse the Google Search result and do not try to deduce the users' search intent. Instead, just provide "null" for the search intent if the keyword is "null".
As output, please provide each keyword and its search intent in the following format:
Keyword 1: keyword 1
Search intent 1: search intent 1
Keyword 2: keyword 2
Search intent 2: search intent 2
Keyword 3: keyword 3
Search intent 3: search intent 3
Keyword 4: keyword 4
Search intent 4: search intent 4
Here is the list of four keywords and their Google search results:
Keyword 1: [[Input fields::Keyword 1]]
Google search result 1: [[SERP Keyword 1::complete_results_text]]
Keyword 2 [[Input fields::Keyword 2]]
Google search result 2: [[SERP Keyword 2::complete_results_text]]
Keyword 3: [[Input fields::Keyword 3]]
Google search result 3: [[SERP Keyword 3::complete_results_text]]
Keyword 4: [[Prompt result::Prompt1]]
Google search result 4: [[SERP Keyword 4::complete_results_text]]

This prompt takes a flexible number of keywords and their corresponding Google search results as input. It instructs the AI system to deduce the search intent for each keyword based on the content of the search result pages, but thereby ignoring the keyword itself. If a keyword is “null”, the AI should not analyse the search result and simply output “null” for the search intent. The output format lists each keyword with its deduced search intent in a clearly defined format (which is necessary for scalability reasons).

Step 3: Get purpose of page

Depending on the website you are using this process for, you might need to adjust this step, as results can vary strongly depending on the content format of the page. Sometimes, I use GPT Vision for this step and instruct it to analyse a screenshot of the page, instead of extracting text content from the page. If the website you are working on has different page types that need different approaches for title tag creation, you might want to include this information in the prompt.

The basic prompt for this step is:

What is the purpose of a web page with the following content? What can a website user find there? Please be as specific as possible and summarise your reply in one sentence. As your output, please provide only the one-sentence summary.
__
[[Website content::main_website_content]]
__

This prompt takes the main content of the web page as input and asks the AI system to determine the purpose of the page and what users can find there. The AI should provide a specific one-sentence summary as output, without any additional explanations.

Step 4: Select the best keyword

Now that we have determined the search intents of our potential target keyword(s) for the page, and we know what the purpose of the page is, it’s time to select the best keyword. We achieve this by matching the search intents with the page purpose to find out which of the search intents is best satisfied by the page content. The corresponding keyword will be the one that we select for the title tag creation.

The prompt for this step is:

You are an internet user who is currently using Google to perform various searches. Please consider the following four keywords and the corresponding search intents, which are your own search intents. Then, compare the description of the page that I provide at the very end of this prompt with each of the four search intents. For which of the four keywords does the described page provide the most satisfying result, considering your search intent for each of the keywords? As an output, please provide only the selected keyword, without any further explanations.
Here are the four keywords and search intents:
[[Prompt result::Prompt2]]
Here is the description of the page:
__
[[Prompt result::Prompt3]]
__

This prompt takes the output of Step 2 (keywords and search intents) and Step 3 (page purpose) as input. It instructs the AI system to act as an internet user performing searches on Google and to compare the page purpose with the search intents for each keyword. The AI should then select the keyword for which the page provides the most satisfying result, considering the search intent. The output is the selected keyword without any further explanations.

Step 5: Create title tag

Depending on the website you are using this process for, you might want to adjust this step. For international websites with different country versions, I normally make sure to include the country name as a variable and instruct the AI system to create a title tag that speaks directly to users from this country.

You can also instruct the AI to use your brand name in a certain format in all title tags, if that’s how you want it to be.

The basic prompt for this step is:

Please write a title tag for the following target keyword:
__
[[Prompt result::Prompt4]]
__
The title tags that you write MUST have the following length: Between 50 and 60 characters (including spaces).
As an output, please provide only the title tag, without any further explanations and without additional characters (such as quotation marks). Do not under any circumstances include the number of characters in the title tag.
The title tag should provide a good description of a page that has the following purpose:
__
[[Prompt result::Prompt3]]
__

This prompt takes the output of Step 4 (selected keyword) and Step 3 (page purpose) as input. It instructs the AI system to write a title tag for the selected keyword, considering the purpose of the page. The title tag should be between 50 and 60 characters long (including spaces), and the output should only include the title tag without any additional explanations or characters.

A note on character length specifications in prompts: LLMs can’t really count, so they will not always manage to generate the correct amount of characters. Previous versions of GPT were quite unreliable when it came to character lengths, but since I switched to GPT-4 Turbo for this type of process, the results are quite satisfactory.

Are you ready for action?

By following these steps and adjusting the provided prompts to your own needs, you can generate optimised title tags that align with the search intent of your target keywords and the purpose of your web pages. Once the process works perfectly for your website, you can use it at scale for bulk optimisations or repeat it whenever you need it.

Your options to get to work are:

• Test this process in PromptMate, by using this public app template.
• Copy this process to another existing software or build your own solution.
• If you don’t have time to adjust this process to your own business’s needs, you can get in contact with me to request a quote for a fully customised version of this process for your website(s).

Fun fact: I want to share as many of my AI processes for marketers as possible with you, but writing guides the old-fashioned way takes up a lot of time. That’s why I’ve also created an AI process for writing guides about AI processes, which has helped me with writing this guide.

Please feel free to share your thoughts in the comments section: