The rapid growth in AI use makes it vital that organisations write in a way that AI can understand. Clear, well-structured text helps AI systems find the right information and present it accurately.

The guidelines below provide 11 principles for writing copy that AI models can easily search and summarise. Each principle includes examples that show how relatively small changes can improve clarity and searchability.

Following these principles should also make your writing easier for humans to follow, which is never a bad thing. One note of caution though: if you follow these guidelines too rigidly, there’s a good chance your words will read like a machine wrote them in the first place. If you need your writing to connect with a real audience, make sure it still feels like one human talking to another,

1 Have a clear structure

AI systems respond best to clearly organised content. Headings, short paragraphs and lists help both people and AI understand a document’s structure.

This approach also improves readability and helps you present key messages more effectively.

Best practices

  • Use headings that describe what the section is about
  • Keep paragraphs relatively short
  • Use bulleted or numbered lists where useful
  • Focus each paragraph on one main idea

Example: before

Our sustainability programme has many parts and has achieved several important successes over the past year, particularly in carbon reduction, waste management, and community engagement.

Example: AI optimised

Sustainability programme overview

  • Carbon reduction initiatives reduced emissions by 18%
  • Waste management improvements increased recycling rates by 27%
  • Community programmes reached 4,200 participants

This structure helps AI systems identify and summarise the key components of the programme.

2 Prioritise key information

AI tools often treat the first sentence in a paragraph as the main point. When you place the most important information first, the system is more likely to capture the right message.

This is called the ‘inverted pyramid’ approach and it’s the standard practice for journalism.

Best practices

  • Open with the key message
  • Put conclusions before supporting detail
  • Avoid hiding the main point later in the paragraph

Example: before

We analysed service desk performance last quarter and found that ticket resolution times have fallen significantly.

Example: AI optimised

Ticket resolution times improved significantly last quarter, decreasing by an average of 22%. A detailed service desk analysis confirmed this change.

3 Use consistent terminology

AI systems treat different terms as different ideas. When you describe the same system or process in several ways, the AI may not recognise that they refer to the same thing.

Consistent terminology also helps people understand more easily.

Best practices

  • Use one clear term for each concept and avoid unnecessary synonyms
  • Define key terms
  • Use the same terminology in headings and body text

Example

Inconsistent terminology

Using client hub, customer portal and online workspace to refer to the same system.

Consistent terminology

Use client portal throughout and clearly define it:

“The client portal is the online platform clients use to access resources, submit requests and track progress.”

4 Use linking words

AI systems rely on language cues to understand relationships between ideas. Linking words and phrases clarify how sentences and concepts connect.

Best practices

Use linking phrases such as:

  • as a result
  • therefore
  • in contrast
  • for example
  • in summary

Example: before

We introduced automated invoicing. Payment times have improved.

Example: AI optimised

We introduced automated invoicing. As a result, payment times improved by 14 days on average.

5 Minimise ambiguity

Vague language reduces the accuracy of AI summaries and search results. Clear, specific language helps both humans and machines understand content.

Best practices

  • Use precise nouns
  • Include measurable details where helpful
  • Avoid vague references like “this,” “things” or “issues”

Example: before

This improved things significantly.

Example: AI optimised

The new onboarding workflow reduced setup time by 35%, significantly improving client satisfaction.

6 Provide summaries and key takeaways

AI systems often rely on brief summaries to understand a section’s meaning. Adding short summaries reinforces key messages for both people and AI.

Best Practices

  • Include executive summaries at the start of documents
  • Add brief summaries at the end of sections
  • Highlight key takeaways where useful
  • Provide glossaries for specialised terms

7 Include structured information

AI systems interpret structured information such as lists, steps and tables very effectively. Structure also helps readers scan and absorb information.

Best practices

Use structured elements such as:

  • step-by-step processes
  • metrics and KPIs
  • bulleted lists

Example: before

The project followed several steps, starting with discovery, then moving into design and implementation.

Example: AI optimised

Project phases

  • Discovery – stakeholder interviews and requirement mapping
  • Design – solution architecture and prototyping
  • Implementation – development, testing, and rollout

8 Use language patterns that match user queries

People often ask questions when using AI search tools. If your content includes phrasing aligned to these questions, AI systems can retrieve it more reliably.

Best practices

Use phrases such as:

  • “This section explains…”
  • “Clients often ask…”
  • “The main reasons are…”
  • “The key outcome is…”

Example

Clients often ask how our onboarding process works. Our onboarding process includes three stages:

  • Orientation
  • System setup
  • Training

9 Reduce non-essential flourishes

Creative language can strengthen storytelling but overly metaphorical phrasing can confuse both readers and AI when clarity matters. AI models are good at understanding common metaphors but you should be avoiding cliches anyway.

Example: before

Our customer support team continues to shine, as the beating heart of our service ecosystem.

Example: AI optimised

The customer support team is central to our service delivery model and resolves more than 2,000 cases per week.

10 Provide self-contained context

AI systems often analyse documents in small chunks, rather than as a whole. Each section must make sense independently.

Best practices

  • Focus each section on one topic
  • Use meaningful headings
  • Avoid very long paragraphs
  • Avoid vague references such as “this process” without context
  • Use clear nouns instead of pronouns where clarity matters

Example: before

This process reduces onboarding time.

Example: AI optimised

The automated onboarding workflow reduces client onboarding time by 35%.

11 Use specific names and details

AI systems retrieve information more accurately when passages contain concrete names, metrics and details.

Best practices

Include:

  • metrics or results
  • timeframes or dates
  • specific names, for example of customers, platforms, divisions or programmes

Example: before

The new system improved case handling.

Example: AI optimised

The ServiceNow case management system improved average case resolution time by 22%.

Find out more

Contact us to discuss how we can help you with your project.