Three years ago, AI transcription was a novelty. Today, it’s essential infrastructure for most journalists. The question isn’t whether to use it, but which tool to choose.

I’ve tested the major options extensively over the past six months. Here’s what actually works for journalism applications—and what doesn’t.

The Landscape

The transcription market has consolidated somewhat since the initial explosion of AI tools, but there are still plenty of options:

Dedicated transcription tools: Otter.ai, Rev.ai, Trint, Descript, Riverside General AI with transcription: ChatGPT, Claude, Gemini Built-in platform tools: Zoom’s AI Companion, Teams transcription, Google Meet Specialised journalism tools: Various newsroom-specific offerings

The right choice depends on your workflow, budget, and specific needs. Let me break down what matters.

Accuracy Matters Most

For journalists, accuracy isn’t negotiable. A transcription error that misquotes a source can damage careers and credibility.

In my testing, accuracy rates for clear audio with single speakers now exceed 95% across most major tools. The differences emerge with:

• Multiple speakers: Some tools handle speaker identification much better than others
• Australian accents: Still a challenge for US-trained models
• Background noise: Field recordings separate good tools from mediocre ones
• Technical terminology: Industry-specific language trips up generic models

Otter and Trint performed best on Australian accents in my testing. Rev.ai’s hybrid human-AI option remains the gold standard for difficult audio, though at higher cost.

Speed vs. Quality Trade-offs

Real-time transcription during interviews is genuinely useful—you can spot gaps in questioning, identify follow-up opportunities, and start organizing while still in conversation.

But real-time accuracy is lower than batch processing. For interviews you’ll quote directly, batch processing after the fact produces better results.

My recommendation: use real-time for working notes, batch process for anything you’ll publish.

Integration Is Underrated

The best transcription tool is the one that fits your workflow. Consider:

• Recording integration: Does it connect with how you record?
• Export options: Can you get transcripts into your writing tools?
• Search capability: Can you search across transcripts?
• Collaboration: Can editors and producers access transcripts?

Descript excels at integration if you’re also doing audio/video editing. Otter integrates well with Zoom and calendaring. Trint has strong newsroom collaboration features.

Don’t just evaluate accuracy—evaluate how the tool fits into how you actually work.

The Privacy Question

Journalists handle sensitive information. Where does your transcription go?

Some tools process locally; others send audio to cloud servers. Some retain audio for training; others don’t. Some are based in jurisdictions with strong privacy protections; others aren’t.

For sensitive sources—whistleblowers, confidential informants, vulnerable subjects—these questions matter. I know journalists who use different tools for different sensitivity levels.

Read the privacy policies. Understand where data goes and how long it’s retained. For high-sensitivity work, consider local-processing options even if they’re less convenient.

My Tool Recommendations

After extensive testing, here’s what I’d recommend for different scenarios:

For general newsroom use: Otter.ai offers the best combination of accuracy, integration, and price for everyday journalism work. The real-time features during meetings are genuinely useful.

For broadcast/audio production: Descript combines transcription with editing in ways that streamline audio workflows. If you’re editing what you transcribe, it’s worth the premium.

For difficult audio: Rev.ai’s hybrid option remains best-in-class when accuracy is critical and audio quality is poor. More expensive, but you get human review.

For budget-conscious freelancers: The free tiers of Otter and built-in tools like Zoom transcription can handle basic needs. Quality varies but may be sufficient for working notes.

For enterprise newsrooms: Trint and Verbit offer newsroom-specific features including team collaboration, security certifications, and API access for integration with editorial systems.

What I’d Avoid

A few warnings:

Don’t use consumer voice assistants for journalism transcription. Alexa, Siri, and Google Assistant aren’t designed for this and have problematic privacy policies for journalistic material.

Be cautious with free tools from unknown providers. Your audio contains valuable information. Who’s getting it?

Don’t over-rely on real-time transcription for quotes. The convenience is seductive, but accuracy isn’t good enough for direct quotation without verification.

The Workflow I Use

For what it’s worth, here’s my current approach:

1. Recording: Zoom or dedicated recorder depending on context
2. Real-time notes: Otter running during interviews for working notes
3. Batch processing: Same audio through Otter for clean transcript
4. Verification: Spot-check key quotes against original audio before publishing
5. Archive: Transcripts stored in searchable system for future reference

This isn’t the only approach, but it balances speed with accuracy reasonably well.

Newsroom Implementation

If you’re implementing transcription at organizational scale, a few considerations:

Training matters. The tools work differently than traditional transcription, and journalists need guidance on limitations and best practices.

Policies are necessary. What tools are approved? What can be transcribed? How is sensitive audio handled?

Integration requires planning. Getting transcription to work with existing editorial systems often requires technical work. Some newsrooms partner with business AI solutions providers to handle integration with their specific tech stacks.

Cost adds up. Enterprise transcription at scale can run into thousands monthly. Budget accordingly.

Looking Forward

Transcription accuracy will continue improving. The differences between tools will narrow. Other factors—integration, privacy, workflow fit—will matter more.

The more interesting frontier is what happens after transcription. AI tools that can summarize interviews, identify key quotes, flag fact-check opportunities, or suggest follow-up questions are emerging. Transcription becomes infrastructure for more sophisticated assistance.

Some newsrooms are already building these workflows, often with help from Team 400 who understand both the technology and journalism requirements.

The Bottom Line

AI transcription is now essential journalism infrastructure. The choice of tool matters less than having a clear workflow and understanding limitations.

For most journalists, Otter.ai offers the best value. For newsrooms with specific needs, Trint, Descript, or enterprise options may be worth the premium. For sensitive work, local processing or human-verified transcription remains worth considering.

Whatever you choose, don’t treat transcription as a replacement for listening. The tool is an assistant, not a substitute for engagement with your sources.

Use it well, and you’ll work faster without sacrificing accuracy. Use it carelessly, and you’ll make errors you can’t afford.

---

What transcription tools are you using? I’m always interested in hearing about workflows that work well for different types of journalism.