How to achieve AI-ready technical documentation

Deploying AI and preparing for AI have become twin imperatives for every organisation. In the context of communicating about complex products, let’s understand these as:

• Deploying AI = using AI to create and manage the technical content of your product
• Preparing for AI = ensuring your product’s technical content is ready to be used by AI

Both are ways of thinking about AI readiness and AI-ready technical documentation. You can make progress with each independently, or treat them as interdependent. 3di has the expertise to accelerate your capability with both.

Avoid unleashing a very confident confusion machine

In both contexts, the fundamental challenge is that AI retrieves information from whatever source material it can access. And, if the source material is a mess, the AI becomes a very confident confusion machine.

When you’re deploying AI, you can be in control of what it’s accessing and fine-tune that source material to get consistently reliable results — that is what 3di has been doing with our range of products that we deploy.

When you’re preparing for AI, you’ve got no control over the AI tool that external readers will be using, and even with internal readers using corporate-enabled AI, you’ve got little control over how they’re going to use it.

With both, we are all also in the dark about what capabilities future-AI will have. Basing a strategy on either it’ll always be rubbish, or it’ll be good enough at everything would seem reckless.

Clear, structured, trusted content is already… AI-ready

The good news, whether you’re deploying, or preparing for, AI, is that there are already proven ways to optimise your technical product content for the GenAI era. Funnily enough, it’s the same ways you can optimise your technical product content for humans: ensure it is clear, structured and trusted. The big exception is visuals, which work well for humans but not, currently, for AI.

3di has been providing this expertise for over 20 years, but in the last 3 years, we’ve invested heavily in adapting that expertise so that our customers’ technical documentation can be AI-ready as well as human-ready.

At the enterprise scale, this collection of capabilities now goes by the name of Content Operations. But even at the scale of a small technology business, 3di both deploys and prepares for AI when we’re delivering our simple Technical Writing services.

Checklist for AI-ready technical documentation

At 3di, we’ve researched, trialled and embedded many of the following approaches. In some cases, they help prepare the content for AI, and in others, we’ve deployed AI to implement or validate the optimum approach as part of the 3di product suite. Here’s an overview:

• Formats: Whether human or AI, PDF’s are a significant barrier, so stop using them online. If you need to print, then no problem, but provide HTML and other text formats too.
• Headings: Headings signal transitions between topics and clarify to AI how content is organised into sections
• Metadata: Page titles and meta descriptions are key signals used by AI systems to understand content’s purpose and scope.
• Tables and lists: Tables and lists organise information into predictable structures so beloved by AI. Maybe not literally, but in a way you are sparing the model one extra processing step.
• Schema: Schema is structured markup that provides explicit, machine-readable context for AI systems to understand what your content represents, not just what it says. The signs are that AI won’t get much value from these over and above writing and structuring well.
• Questions and answers: Structuring content into the question-answer format mimics how people search for information. AI assistants can often reuse such structures verbatim in generated responses.
• Punctuation: Avoid decorative punctuation like ellipses, excessive em-dashes, or multiple exclamation marks. Stylised language can confuse AI.
• Style: Clarity and consistency in writing make it easier to interpret your content accurately and reuse it in meaningful ways.
• Visuals: Visuals are not a priority in AI processing by default – regular text comes first. Analysing images is expensive and not very reliable. If you must use visuals, make sure you add context around them: HTML alt attributes, figure titles, and descriptive text that reflects the ideas expressed in the visual.

With 3di’s expertise, you can ensure your technical documentation is AI-ready.