Schema markup is arguably the best way to provide more context to search engines about the content of your website and its pages.
It’s essentially code that describes code.
One of my favourite use cases of ChatGPT is to help it generate, debug and scale schema markup.
I want to share some useful prompts that will help you achieve this.
ChatGPT prompts for schema markup
Make sure that when you generate schema markup with ChatGPT, that you don’t forget to include the opening and closing script tags that are necessary if you want to implement your schema markup.
Opening tag: <script type=”application/ld+json”>
Closing tag: </script>
With that out of the way, here are some useful ChatGPT prompts for schema markup:
1. Generating ideas for schema types
This is ideal if you are starting out and aren’t sure what schema types you should be using for your website. You can get away with being a bit more generic here, but it doesn’t hurt to specify exactly what pages you want the schema markup on, as well as what you want it to achieve e.g. to demonstrate that we have an offer for a limited time. Here is a prompt template to use:
I want to create JSON-LD schema markup for my B2C/B2B website, X (URL) that provides X. Provide me with some specific schema markup types that is relevant to my industry and what my offerings are. Be sure to look into the ‘More specific Types’ that Schema.org suggests to find the one most applicable to my situation.
2. Generating schema using predetermined values
This works best when you’ve found a schema type you want to use and have noted down the properties and values that you want to include. At this point you should be very specific with the prompt and outline exactly what you want included. Here is an example prompt for generating Recipe schema:
Generate Recipe schema markup in JSON-LD using the following properties and their values:
name: Chocolate chip cookies
url: https://www.itamarblauer.com/cookies-recipe
author: Itamar Blauer
sameAs: https://www.linkedin.com/in/itamarblauer/
cookTime: 1 hour
cookingMethod: Baking
recipeIngredient: Chocolate and stuff
about: chocolate chip cookies
sameAs: https://www.wikidata.org/wiki/Q14169302 and https://en.wikipedia.org/wiki/Chocolate_chip_cookie
Don’t forget to include the opening and closing JSON-LD script tags
This prompt worked and displayed 0 errors.
3. Generating ideas for additional properties to include
If you’re not fully sure that you have actually used all the relevant properties for your case, after you get the output from the above prompt, you can ask ChatGPT the following:
What additional properties would you recommend including in addition to the schema you have just displayed?
4. Fixing schema
If you did some tinkering in the Schema Markup validator tool and saw there are errors, you can try this prompt:
Fix this schema:
*paste schema here*
Then, check the new output with the validator tool to see if it worked (which it usually does).
5. Scaling schema
To scale schema markup, you will require a .xlsx spreadsheet with multiple tabs that contains the properties in the first column and the values in the second column.
You will also need the premium version of ChatGPT (GPT-4) and use the Advanced Data Analysis option so you can insert the spreadsheet file.
Once you uploaded the file, tailor the following prompt to your needs (this is an example for multiple Recipe schemas):
There are multiple tabs in this file. The first column contains the properties and the second column contains their values. Generate Recipe schema separately for each of the tabs using JSON-LD making sure to include the JSON-LD opening and closing script tags. Only use the data in the tabs and do not add any other properties. Make sure each tab has its own schema generated with the referenced properties and values.
When doing this at scale, you have to be certain that the schema is correct, which in the example above, isn’t exactly how I would have expected based on the contents in the spreadsheet.
The most fool-proof way to get this right is to start off with one tab in a spreadsheet that contains all the info you need, and make sure that the Advanced Data Analysis feature within ChatGPT can correctly display the schema.
If it doesn’t, tweak it, and once it does, scale up the amount of tabs within the spreadsheet.
Remember to always check your schema with the validator tool before implementing it!
Deck: Generating complex schema markup at scale with AI
Here is the deck related to this topic that I delivered at brightonSEO on 14th September 2023:
Conclusion
The use of ChatGPT and AI can help with the generation, troubleshooting and scaling of schema markup.
However, the most important concept to remember is GIGO (garbage in, garbage out).
If you prompt badly, you’ll not get your desired output.
Have fun experimenting with these prompts, and if you need additional SEO help, you can always count on an SEO consultant!
