Main content

Contentstack integrates ChatGPT into headless CMS editing environment

Barb Mosher Zinck Profile picture for user barb.mosher April 12, 2023
Generative AI, meet the CMS!


Is your content management system integrating with ChatGPT? Unless you are using another generative AI tool/writing assistant, then it’s something you might want to ask your CMS provider about. If you are a customer of Contentstack, then you know the answer.

Contentstack announced its integration with OpenAI’s ChatGPT 3.5 Turbo in its headless CMS, saying it wants to help companies accelerate content creation. Currently in testing through its early access technology program, content creators and editors can create, test, and translate short and long-form content in the entry editing experience.

The AI assistant

No cutting and pasting here. The AI Assistant is available in-line for every text-based field in the editor. So you can use it to create summaries, titles, descriptions, tags, headlines, and even outline or write complete blogs. Contentstack also integrates ChatGPT to provide translations for text with a single prompt. 

You can create content using prompts, as well as request rewrites and text shortening. Or you can generate content based on the content in another field you specify. 

The AI Assistant is not implemented by default but is available through the Marketplace. Conor Egan, VP of Product and Engineering, said Contentstack understands that not every company is ready for generative AI, so they provide it as an option and have made an effort to help educate customers on using AI to help with their content creation processes.

Egan said the company has been experimenting with AI for a while. One of the areas they looked at was implementing AI in their automation hub, which offers automated workflows. But concerns with AI hallucinations and incorrect information led them to determine that, for now, it was better to include the capabilities in the context of supporting people. Using this approach, content creators and editors can choose when to use it and whether or not to accept it. 

Another important option that is available with v3.5 is to specify that your data does not go back into the model. In Contentstack, this option is enabled by default, which is good, because Egan said that they have customers coming with requirements in their contract that none of their data be used to help train the model.

According to Egan, a number of customers are showing interest in the AI Assistant. The use case he’s seeing the most is for summarization rather than writing from scratch. Egan noted that most of these companies have writers, so the goal is more about helping them be more productive than replacing them. However, he believes that may change as AI gets better at understanding a brand’s voice. 

Just getting started

Contentstack is always evolving, and how it decides to use AI in its platform will also evolve. Egan said they are starting with out-of-the-box use cases like generating text and images. For example, he said they are looking at having the AI read a blog and generate five images to go with it for the content creator to choose from. 

As AI grows more trustworthy, we’ll see more automation. For Contentstack this means helping model content (e.g., ingest a website and recommend a content architecture) or create customer-specific models. Or getting back to the automation workflows, they are looking at how they can use AI to tailor the user experience, including offering recommended actions or performing actions automatically. 

Egan said that every person who uses a CMS shouldn’t have the same experience. The person from legal who only goes in to review and approve content should have a different experience from the content writer who spends the better part of their day in the CMS. Understanding how they can tailor the entire experience in the software is part of their future plans.

One final area where AI and can support better experiences comes from Contentstack’s new hosting product. Having the front-end and back-end close together gives Contentstack more data on how content is consumed. The company will be able to tie that back to what’s happening on the back end, including recommending new content or changes to existing content. 

My take

As a marketer and one who writes about marketing technology, I have chosen not to jump wildly on the generative AI bandwagon but to try to learn and understand as much as I can about how marketing is and will evolve because of it. 

Gartner believes that by 2025, 30% of all marketing content will be generated by AI. And it’s very likely this will be true. My hope is that organizations will think carefully about how they choose to use it, especially as a tool to help both marketers and their writers be more productive. That’s why I like Contentstack’s decision to integrate ChatGPT as an enabler. I expect that many other CMS platforms are in the process of their own integrations. 

Tooba Durraze, VP of Product (AI and Data) at Qualified had this to say about the CMS market: 

We're going to see an influx in a new style of CMS, with it pre-generating and then marketers approving text before it ends up anywhere on their websites. The middle ground between generating text and then using it as output is where the innovation is going to come from. And in those cases, the master connectors -- or people who connect to a ton of integrations, people who have a strong data pipeline and data foundation -- are the people that are going to come out on top.

She’s right. But there’s more that’s going to change in marketing than content management systems. When products like Zapier integrate with ChatGPT and HubSpot’s start to really take off, we all need to hold on to our seats. 

A grey colored placeholder image