The short version: A typical transcription job means listening to audio in short bursts, typing at speed, formatting to a strict style guide, and editing your own work against a deadline, usually for somewhere between 15 pence and 1 pound per audio minute depending on difficulty. Most days involve far more editing and quality checking than actual typing from scratch, because AI tools now do the first pass on easy, clear audio and humans get called in to fix what the machine got wrong or handle the audio that’s too messy for it. If you’re picturing quiet, easy background income, the reality is closer to piecework in a call centre, just done from your kitchen table.
What a real transcription day looks like, hour by hour
I hired freelance transcriptionists for close to a decade to turn my keynote recordings, webinars and podcast interviews into blog posts and show notes, so I’ve watched this work from the client side more than most people writing about it. One woman I worked with regularly, based in Manchester, used to log her days for me because I paid per finished word, not per hour, and she wanted to prove her rate was fair. Her Tuesday looked like this:
- 8.00am to 8.45am: reviews the audio file, checks the length (my talks ran 40 to 55 minutes), listens to the first two minutes to catch accents, crosstalk, or bad mic quality before quoting a turnaround time
- 8.45am to 11.30am: first pass typing, using foot pedal software to slow playback and rewind without touching the keyboard
- 11.30am to 12.15pm: break, because typing at speed for three hours straight gives you the same wrist and neck strain as long-haul data entry
- 12.15pm to 2.00pm: second pass, checking her own transcript against the audio for missed words, wrong names (my clients’ names were never in any dictionary), and timestamp accuracy
- 2.00pm to 3.00pm: formatting to my style guide, speaker labels, paragraph breaks every 200 to 300 words, removing filler words unless I’d asked for verbatim
- 3.00pm to 3.30pm: proofreading against a checklist, then delivery
That’s roughly six and a half working hours to produce a 45 minute file, and she billed me around 65 pounds for it. Do that maths against an hourly rate and it’s under 10 pounds an hour once you account for the review and formatting time nobody quotes you when they describe “just typing what you hear.”
The actual daily tasks, broken down
Transcription work is not one task, it’s five, and the ratio between them is the thing nobody tells you upfront.
- Listening and typing: the smallest part of most days now. Clear, single-speaker audio in a quiet room is exactly what AI tools like Otter, Rev’s automated engine, and Whisper handle well, so a lot of paid human work has shifted away from this.
- Editing AI output: increasingly the main job. You get a rough auto-generated transcript full of wrong names, garbled technical terms, and misheard homophones, and your task is to fix it against the audio, which is slower than people expect because you still have to listen to almost the whole file to catch the errors.
- Formatting: every client or platform has a style guide. Legal transcription wants verbatim including “um” and false starts. Corporate content wants filler removed and grammar cleaned up. Getting this wrong means rejected work and no pay.
- Research: looking up proper nouns, company names, medical or technical terminology mentioned in passing, product names, place names. A single unfamiliar surname can cost you five minutes.
- Quality control and resubmission: on platforms like Rev or GoTranscript, files get graded and sent back for revision if accuracy drops below a set threshold, often 98 to 99 percent, and revisions are unpaid.
The pay reality nobody puts in the recruitment ad
Here’s the part most “start a transcription side hustle” posts skip past. Ten years ago, general transcription on freelance platforms paid roughly 0.60 to 1.10 pounds per audio minute. In 2026, rates on the big platforms have compressed, often sitting between 0.20 and 0.55 pounds per audio minute for general work, because AI-assisted first drafts mean companies pay humans for editing time rather than typing time, and editing time is billed lower. Specialist transcription, legal depositions, medical dictation (handled by trained medical transcriptionists working under strict confidentiality, not something to attempt without training), or heavily accented multi-speaker audio, still commands 1 to 2 pounds per minute or more, because that’s the audio AI still struggles with.
A realistic new transcriptionist, working consistently, might clear 400 to 700 pounds a month part-time, or 1,200 to 1,800 pounds a month full-time once they’re fast and accurate. The people earning more than that have usually specialised, in legal, academic research interviews, or captioning for broadcast, where the audio is harder and the client pool is smaller.
What separates someone who lasts from someone who quits in six weeks
Typing speed matters less than people assume. What predicts whether someone sticks with transcription work is:
- Tolerance for repetitive, isolating work with no colleagues to talk to during the day
- Comfort using a foot pedal and playback software rather than just headphones and a pause key, which roughly doubles your speed once you’re used to it
- Willingness to build a personal glossary of terms for repeat clients, so you’re not re-researching the same names every file
- An ear for accents and audio quality, since a large chunk of daily income depends on which files you get assigned, and difficult files pay more but take longer
- Discipline around posture and breaks, because repetitive strain injury is the single biggest reason experienced transcriptionists leave the field
The uncomfortable part
The transcription job market of 2026 is not the same job that existed in 2015, and pretending otherwise does new entrants a disservice. AI transcription accuracy on clean audio is now high, well into the 90s percent for a single clear speaker in English, which means the low-skill entry point, sitting down and typing what you hear from an easy file, is mostly gone. What’s left is harder: editing machine output against messy real-world audio, working on files with several overlapping speakers, strong accents, poor recording quality, or specialist terminology. That’s a smaller, more skilled pool of work, and it pays better per file but there’s less of it to go around. Anyone selling transcription as easy, beginner-friendly passive income right now is describing a job that mostly doesn’t exist anymore. The people still earning a decent living at it retrained around AI rather than against it, learning to use tools like Otter or Descript as a first pass and positioning themselves as the accuracy layer on top, not as typists competing with software that types faster and never gets tired.
That shift mirrors something I’ve written about a lot in my own work rebuilding my business around AI rather than resisting it. If you want a practical route into understanding how these tools change a job rather than just threaten it, I’d point you to how to become fluent in AI in 90 days, because the same logic that applies to marketing applies here: the people who survive are the ones who learn to sit on top of the tool, not underneath it.
How to get started, if you still want to
If you’ve read all that and it still sounds like your kind of work, here’s the real path, not the vague one:
- Take a free typing test to confirm you’re at 60 words per minute or above with 98 percent accuracy, most platforms require this before they’ll even test you
- Apply to two or three platforms at once, GoTranscript, Rev, and TranscribeMe are the common entry points, and each has its own paid or unpaid assessment file
- Buy a foot pedal, a decent pair of closed-back headphones, and transcription software with variable speed playback before your first paid file, not after
- Track your actual minutes worked against pay for your first month so you know your real hourly rate, not the advertised one
- Specialise within six months, pick legal, academic, or a technical niche you already understand, because generalist rates keep falling and specialist ones don’t
Keeping half an eye on how AI tools are moving is worth the ten minutes a week it takes, and roundups like AI news this week for small business are a useful way to see which tasks are being automated next before it catches your niche by surprise.
I have written more around this on the site: AI News This Week for Small Business, 26 July 2026, How to Become Fluent in AI in 90 Days (Even If You’ve Already Tried and Given Up).
Frequently asked questions
Do transcription jobs pay well in 2026?
General transcription pays modestly, often 0.20 to 0.55 pounds per audio minute on the main platforms, working out to roughly 400 to 1,800 pounds a month depending on hours and speed, while specialist transcription in legal or academic work can pay two to four times that because the audio is harder for AI to handle.
Is transcription mostly typing or mostly editing now?
For most paid work in 2026 it’s mostly editing. AI tools handle a rough first pass on clear audio, and human transcriptionists are increasingly paid to correct names, technical terms, and errors against the recording rather than type from a blank page.
Can I do transcription work with no experience?
Yes, but expect an unpaid or low-paid assessment file first, and expect the entry-level general work to pay less than it did five years ago, since that’s exactly the tier AI has moved into.
What equipment do I need for daily transcription work?
A foot pedal, closed-back headphones, and transcription software with adjustable playback speed are the three things that make a real difference to your daily output, along with a quiet room and a second monitor if you can manage it.
