OpenAI has just announced ChatGPT Atlas, a web browser with AI built in. It comes hot on the heels of other AI browsers, such as Dia and Comet. And Anthropic hasn’t been idle – it’s recently released a desktop app version of Claude.
The question is: How will these technologies reshape our workflows, and what does this mean for our careers?
ChatGPT Atlas
ChatGPT Atlas represents OpenAI’s entry into the browser market, directly challenging Google Chrome. Atlas places ChatGPT at the core of the browsing experience, rather than as an add-on feature.
The key differentiators for technical writers include:
AI integrates into more of your daily activities
An “Ask ChatGPT” button appears on every webpage, opening a sidebar that automatically understands what you’re viewing. If you’re researching something like API documentation or reviewing competitor docs, this eliminates the need to copy-paste between a browser and a chatbot.
Better contextual memory
Atlas can remember websites you visit and what you do on them. It provides more responses based on your browsing history. If you’re looking for patterns, it can synthesise that information automatically from your browsing history. You won’t be relying on having to manually compile notes or using a web clipping app.
Agent mode on hand
ChatGPT Agent mode is an existing feature on the paid versions of ChatGPT. It’s an AI agent that can autonomously do research and carry out tasks such as navigate websites, fill forms, and complete multi-step tasks. In the future, we might start with a query in the browser and then ask ChatGPT Agent to take over and complete the task. This hints at the research stage of technical writing becoming partially automated.
Comet
Comet is another AI browser from Perplexity. The key differentiators for technical writers include:
Threads and branching
Comet can branch into subtopics without losing original context. In a technical writing context, it might prove valuable when documenting complex software systems with interconnected components. You could explore different branches in your research and use AI to help synthesise the information.
This also a feature of Claude Projects (which isn’t a browser but a new part of Claude).
Working in the background
The paid version of Comet can carry out multiple tasks simultaneously in the background. for example, it could be benchmarking your existing documentation against competitor documentation, while you continue writing.
Claude’s new features
Let’s move onto the new Claude-related features. The key differentiators for technical writers include:
Direct file access
While not a browser, Claude Desktop (with its command-line tool Claude Code) can read and edit files in your project directory, analyse Git branches, and maintain context about your documentation structure.
For example, Mintlify’s documentation team is using Claude Code to document new features by having it analyse code changes and draft initial documentation based on actual implementation.
Project context understanding
Anthropic is improving its project context understanding capabilities with CLAUDE.md files. Through these files, you can train Claude to work how you do. You can teach it rules about your documentation style, terminology preferences, and quality standards. It now has the intelligence to know when to bring that knowledge into its memory. This should lead to better consistency in the chatbot, and consequently across documentation updates.
Verification capabilities
Claude Code can test documentation links, verify code examples against actual codebases, and cite sources properly. This might save time in a technical writing project.
Where is this all heading?
In the round, these tools could enhance the technical writing process.
Accelerated information gathering
In addition to trying to sit down with a Subject Matter Expert, traditional technical writing research often involves desktop research: opening dozens of tabs, manually comparing information sources, and synthesising findings into documentation.
AI browsers offer the promise of:
- Pattern recognition at scale
- Real-time fact checking
- AI browsers with access to current web information can verify technical claims, check for updated versions, and flag potential inaccuracies as you write.
Claude Code’s ability to read codebases and analyse Git branches addresses a persistent technical writing challenge: getting timely information from busy engineers.
Automated quality checking
AI is good at repetitive tasks that don’t require deep expertise:
- Checking link validity across documentation sites
- Comparing feature implementations across product versions
- Creating initial drafts based on code or UI changes
- Maintaining consistent terminology across large documentation sets
A new documentation development lifecycle
A typical technical writing workflow for an API is:
- Meet with SMEs to understand feature
- Review code/specs manually
- Draft documentation
- Multiple review rounds with SMEs
- Final editing and publishing
We could be heading towards a workflow that instead looks like this:
- Meet with SMEs to understand feature and user needs
- Use Claude Code or AI browser to analyse the implementation
- Generate initial draft with AI assistance
- SME reviews drafts for accuracy and edge cases (not basic drafting)
- Refine for clarity, user focus, and information architecture
- Publish
You’ll notice steps 2 and 3 change. In an ideal world, the time saved there could be reinvested in step 1 (deeper user research) and/or step 5 (higher-quality refinement).
There are risks
Security and privacy implications
Browser memory features and AI agents that see everything you’re viewing raise significant concerns:
- The AI lethal trifecta can enable baddies to gain access to sensitive data.
- Technical writers often work with pre-release product information, internal APIs, and other confidential information. Browsers that remember what you view and send that context to AI models create potential security risks.
- Many technical writers work under NDAs or for organisations with strict security policies. The broad permissions required by AI browsers, such as Comet’s extensive Google Account access could violate client agreements or corporate security policies.
It might be safer to wait until these concerns have been addressed before using an AI browser for work involving confidential information. A lot of the capabilities in AI browsers are available in other AI tools with more controlled access.
The AI career-eating leopard?
The most significant career risk is that AI browsers and assistants could commoditise basic technical writing tasks. If anyone with Atlas or Comet can generate adequate documentation or straightforward procedures, what distinguishes professional technical writers?
Early evidence suggests there’s still a need for a human-in-the-loop:
- AI excels at structure, but struggles with substance
- AI tools can generate well-formatted documentation that looks professional but often contains subtle technical inaccuracies, misses important edge cases, or fails to address actual user pain points.
- Context and judgment remain human domains
- Understanding which information users need, how to sequence complex procedures, and when to provide additional context requires judgment that current AI lacks.
The skill atrophy risk
As AI handles more routine writing tasks, there’s a risk that newer technical writers won’t develop foundational skills:
- If AI browsers always synthesise information, writers might lose the ability to critically evaluate sources, identify authoritative information, and detect bias or inaccuracy in technical content.
- Relying too heavily on AI to bridge gaps in technical understanding could prevent technical writers from developing the deep expertise that makes them valuable long-term collaborators with engineering teams.
- And there’s also the risk of losing institutional knowledge. If teams don’t document their documentation standards, style guides, and quality criteria in accessible formats, this expertise could be lost when experienced writers leave or when teams rely too heavily on AI defaults.
Quality and accuracy concerns
AI hallucination is a feature not a bug.
- AI-generated technical content often sounds authoritative and uses correct terminology while being factually wrong.
- This is particularly dangerous in regulated environments.
- It’s also a big problem in software documentation where incorrect procedures can cause data loss or security vulnerabilities.
- It could erode user’s trust in the documentation. Once users learn they can’t rely on your docs, they’ll stop consulting them.
What you save in time in one place, you lose in fixing what AI has created:
- AI comes with a verification overhead
- The time required to thoroughly verify AI-generated content can sometimes exceed the time to write documentation manually.
The persistent context challenge
AI browsers with memory have a downside – they can introduce bias from past projects. They might also contain out-of-date information.
Technical writers need to develop practices for:
- Regularly reviewing and pruning browser memories
- Starting fresh contexts for new projects
- Recognising when AI suggestions reflect outdated project information
Career strategy: Positioning yourself for the future
Positioning your role
The challenge is to be seen as a person with a tool that augments their capabilities (giving them superpowers), not a person who can be replaced by a tool:
- Leading and advocating
- Knowing when and where to correct a system (quality gatekeeper)
- Knowing how to configure
- Knowing and preventing the risks (control owner)
- Focusing on high-value activities (planning and directing)
- Building your unique value
Developing new skills
Cherryleaf’s Managing and mastering documentation projects with AI instructor-led course enables people managing a documentation project to use AI to manage the project more effectively.
And our Using Generative AI in technical writing e-learning course teaches technical communicators how to use generative AI tools like ChatGPT and Claude to improve their efficiency and deliver better technical documentation.
Summary
AI browsers probably do represent a fundamental shift in how technical writers work, but not necessarily in whether technical writers remain valuable. The technical writers most at risk are those who see their role primarily as “writing down what engineers tell me”, or “making documentation look professional.” These tasks could end up being automatable.
While the tools are changing, the fundamental value proposition isn’t : helping users successfully use complex software. The question isn’t whether to adopt these AI tools but how to adopt them strategically while preserving and highlighting the uniquely human value you bring to technical documentation.
Leave a Reply