Cherryleaf’s 2026 survey of AI in technical communication reveals technical communicators are no longer asking whether AI has a place in documentation work. Most are already using it.
In our 2026 survey, 62% of the respondents said they use AI regularly or daily in their role. Only 8% said they do not use it at all.
That is a clear shift from previous years. AI is now part of the documentation workflow for many teams.
The harder question is whether those workflows are mature enough. At the moment, the answer seems to be, not yet.
The tldr;
The tldr:
- 109 people responded. Most are technical writers/authors, largely in software/tech, with a strong UK/US weighting.
- AI use is now mainstream
- Confidence in AI is moderate, not expert
- People are still mostly self-taught
- Agent readiness is still early
- Automation is emerging but not universal
AI is being used for practical documentation work
Respondents are using AI for familiar, everyday tasks:
- Drafting and rewriting content
- Improving tone and clarity
- Summarising source material
- Creating outlines
- Reviewing documentation
- Checking consistency
- Working with code, API docs, YAML, JSON, and examples
- Turning support tickets, Slack threads, Jira issues, and SME notes into usable source material
Rather than showing a picture of AI replacing technical writers, the survey suggests AI being used to handle rough material, speed up first passes, and reduce some of the friction around repetitive documentation work.
Of course, the survey might have attracted only respondents who are more AI savvy than other Technical Authors.
The strongest reported benefit was speed. People talked about faster release notes, quicker first drafts, faster internal training materials, and less stress around routine updates.
AI can move the work, rather than remove it
But there was a catch. Several respondents described a new problem:
- AI makes it easier for other people to generate documentation drafts, but those drafts still need expert review.
Sometimes they need a lot of review.
One respondent described the risk of technical writers becoming overwhelmed by low-quality pull requests from colleagues using AI.
Another pointed out that a weak AI-generated draft can take longer to fix than a human-written draft.
AI can reduce the effort needed to produce text.
It does not automatically reduce the effort needed to produce accurate, useful, maintained documentation.
For many organisations, the bottleneck could move from writing to reviewing.
Documentation is becoming source material for AI systems
39% of respondents said their organisation already has an AI system, such as a chatbot, that draws on user documentation. Another 36% said one is planned or in development.
That changes the role of documentation.
A help topic becomes source material for an AI assistant answering customer questions, support queries, or internal product questions.
If the documentation is unclear, outdated, incomplete, or inconsistent, the AI system can repeat those problems at scale.
For technical communicators, that creates a new responsibility: making documentation reliable enough for both humans and machines.
Most people are self-taught
The survey shows a profession learning by doing.
76% of respondents said they had not received formal training in using AI for technical writing. Most are self-taught.
Why does that matter?
Opening ChatGPT, Claude, Gemini, Copilot, or another tool and asking it to improve a paragraph is easy.
The harder part is knowing:
- When AI output is good enough
- When it is quietly wrong
- How to give it useful source material
- How to review what it produces
- How to stop poor drafts increasing the workload for technical writers
- How to prepare documentation for AI assistants and agents
They are documentation management questions. And people might not be able to acquire them without some form of training.
AI-agent-ready documentation is still in its infancy
Only 27% of respondents said they had adapted their content or documentation site to be AI-agent-friendly. Another 29% plan to do so within six months.
That leaves 44% who have not.
This is one of the most interesting gaps in the survey. People are using AI heavily, but many documentation sites and content sets have not yet been prepared for AI use.
AI-agent-friendly documentation might include clearer structure, better metadata, semantic markup, machine-readable formats, controlled terminology, stronger information architecture, and files such as llms.txt.
Although the exact approach will vary, the principle remains the same:
- if you want AI systems to retrieve and use your documentation reliably, the content has to be structured and maintained with that use in mind.
The role of the technical communicator is changing
The survey responses show concern about the future of the role. We got a sense of a practical unease, rather than confidence or panic.
If AI can produce drafts, what should technical writers own?
The answer may be less about typing the first version and more about:
- Defining what good documentation looks like
- Designing content structures
- Checking accuracy
- Managing source material
- Improving review workflows
- Preparing content for ai retrieval
- Setting standards for ai-generated drafts
- Making sure the final answer is useful to the user
As a consequence, AI raises the importance of human judgement. Someone still has to know whether the answer is correct, whether the explanation fits the user’s situation, and whether the documentation can be trusted six months later.
What organisations should do next
The survey indicates AI adoption has moved faster than AI working practices.
Technical communication teams now need to define how AI should be used in documentation work.
That includes decisions such as:
- Which tools are approved
- What information can be put into ai tools
- Which tasks ai can support
- Which tasks need human ownership
- How ai-generated drafts are reviewed
- How quality is measured
- How documentation should be structured for ai assistants and agents
Without that, AI use stays informal.
Some people get faster. Some produce weak drafts. Technical writers inherit more review work. The organisation gets no clear measure of quality or risk.
The practical opportunity
For technical communicators, this is a good moment to take the lead.
Documentation teams understand structure, accuracy, audience needs, terminology, review processes, and maintenance. Those are exactly the things AI systems need if they are going to answer questions reliably.
The tools can draft text, summarise source material, and suggest structure.
They do not remove the need for someone who understands the product, the user, the source material, and the standard the answer must meet.
That is where technical communicators still matter. In many organisations, they could matter more than before.
Note: We used Codex to analyse the data and NotebookLM to generate the images in this post.
Leave a Reply