DavidImagine waking up to a fully edited YouTube video, with every cut, sound effect, and animation already in place.
That is exactly what YouTuber Christian Peverelli does with his AI video editing system built on Claude Code, and he says he doesn’t know how to code or edit.
Continue reading, to find out how to automate video editing with Claude Code, the tools you need, how to set everything up, and exactly what it costs per video.
Everything below comes from his full walkthrough on YouTube.
- What Is the Claude Code Video Editing System?
- Job 1: Claude Code as the Editor
- Job 2: Parakeet as the Transcriber
- Job 3: Hyperframes as the Animator
- Job 4: Google Chrome as the Researcher
- Job 5: Tella as the Screen Editor
- Job 6: Epidemic Sound as the Sound Designer
- Job 7: FFmpeg as the Finisher
- What Is a Claude Code Skill?
- How to Set Up Claude Code for Video Editing
- How to Connect Your Video Editing Tools
- How to Teach Claude Your Editing Style
- Claude Code Video Editing in Action
- How Much Does Claude Code Video Editing Cost?
- Why Automating Video Editing Saves More Than Money
What Is the Claude Code Video Editing System?
Making a video used to take a small team. Christian still has one job, which is to hit record and give feedback.
Every other job goes to a different tool, and Claude Code coordinates all of them.
In total, the system covers seven jobs: editor, transcriber, animator, researcher, screen editor, sound designer, and finisher.
Here is how each one works.
Job 1: Claude Code as the Editor
A human editor sits in front of a timeline for hours. They watch every clip, keep the good takes, cut the bad ones, and decide where every zoom, graphic, and animation goes.
In this system, Claude Code does all of that. Think of it as the project manager of the entire video.
It uses the desktop version of Claude, where Claude Code can work locally. That means it uses your computer, accesses your files, runs other programs, and saves the finished output to a folder.
There is one catch. Claude can’t hear your videos, so it needs another tool to know what you said and where you made mistakes.
Job 2: Parakeet as the Transcriber
A transcriber listens to all the footage, types it out, and adds timestamps. In this system, that tool is Parakeet.
Parakeet is completely free and was built by Nvidia, the AI chip maker.
Once Parakeet produces the transcript, Claude reads the timestamps and the exact words spoken.
That is how it identifies the right takes and chops up the footage.
Job 3: Hyperframes as the Animator
The animator builds every title, logo pop-in, and motion graphic. In the old world, this meant working in After Effects.
Here, Hyperframes does the job. It is a free tool from HeyGen, the video platform.
Claude writes each animation like a tiny website, and Hyperframes turns it into a 4K video. The animations stay sharp even on a large TV.
Everything runs on your own computer, so it costs nothing.
Job 4: Google Chrome as the Researcher
Sometimes Christian says things like “OpenAI just announced a new feature.” When he does, the video needs proof on screen.
That is the researcher’s job. They find a screenshot of the announcement, a post on X, or a research paper, and turn it into B-roll, which is outside footage shown on top of the video.
The tool for this is one you already have, which is your browser. Christian uses Google Chrome, and Claude uses it to do the research.
Claude can grab screenshots, record video of a page scrolling, highlight lines, and zoom into specific sections.
It only shows exact facts, taken from the original post or web page. So he can say what happened and the proof shows up on screen.
Job 5: Tella as the Screen Editor
Christian makes a lot of tutorials that teach people how to use AI tools and make money with them, so screen sharing is a big part of his videos.
For that, he uses Tella, which records your face and your screen at the same time.
Tella has a connector called an MCP, so you can plug it straight into Claude Code.
Anything you can do inside Tella, such as cuts, zooms, layouts, and formatting, Claude Code can do for you. You only record, and it does the rest.
Job 6: Epidemic Sound as the Sound Designer
Every good video team has a sound designer. They choose the music for each clip, add sound effects when things happen, and keep it all on brand.
You probably don’t want music that makes viewers sleepy in the middle of a tutorial.
Christian uses Epidemic Sound, a paid library of music and sound effects, and connects it to Claude.
Claude searches the library for sounds, and you can train it on what you like.
His preference is a slightly suspenseful hook, followed by upbeat lo-fi hip-hop music kept very low in the mix. Sound effects get their own rules too.
Job 7: FFmpeg as the Finisher
The finisher joins all the pieces together and checks for mistakes. That tool is FFmpeg.
FFmpeg is free and has been cutting videos for more than 20 years. It sits under the hood of pretty much every editing program that exists today.
Claude uses it to cut out black screens and check that the audio is in sync, before Christian ever sees the video.
What Is a Claude Code Skill?
None of those seven tools know what your video is supposed to look like. That is the job of the skill.
A skill is a file of instructions that Claude reads before it does the work. It is your editing style broken down into specifics, such as colors, fonts, animations, sounds, and personal preferences.
Christian’s skill is called YouTube-Edit and runs about 1,400 lines. Claude Code built the whole thing for him.
In short, Claude runs the tools, and you tell Claude what good looks like.
How to Set Up Claude Code for Video Editing
You don’t need to be technical to set this up. Follow these steps:
- Download Claude. Get the desktop app from Anthropic’s website and create an account with Google or a username and password.
- Upgrade to a paid plan. You need one for this. Christian recommends the $20 plan, and there is no need to go higher at this point.
- Open Claude Code. In the app, start a new session and switch from the normal Claude view (chat and Cowork) to the Claude Code tab.
- Select Local. This makes Claude use your own computer.
- Choose a working folder. Pick an existing folder or create a new one and attach it.
- Set permissions to auto. This stops Claude from asking you to approve thousands of small actions.
- Pick one of the latest models for setup. Setup is the important step. Afterward, the system can run on a cheaper model.
- Adjust the Claude Code settings. Open settings, go to Claude Code, and turn on the option that lets Claude browse the internet. You can also copy Christian’s exact settings.
Keep in mind that the interface changes often, so it may look slightly different from what you see in the video.
How to Connect Your Video Editing Tools
Claude Code is good at setting itself up, so you can simply ask it to build the system for you.
Christian’s prompt was: “Could you please create a video editing system for me? I’d like to use Hyperframes, FFmpeg, Tella, and Epidemic Sound for this system. Run me through how to set it up step by step.”
Claude can download and install most of what it needs. A couple of tools need to be connected by you:
- Tella and Epidemic Sound: On the left, go to Customize, then Connectors. Search for the tool and click the plus button if it isn’t attached. Epidemic Sound asks you to log in through a separate browser window, and a check mark appears when it’s done.
- Hyperframes: If Claude can’t find it, search for Hyperframes to find the repository link. You can also find it inside the connectors.
After everything is connected, ask Claude to save the setup as a skill on your computer. Then open a new thread and give it a first test.
Setting up is the easy part. What takes time is fine-tuning the look and feel.
If you’d rather skip that, Christian shares his skill for free in his WeAreNoCode community.
How to Teach Claude Your Editing Style
Christian started small. He gave Claude notes on everything he made, beginning with short clips, usually just the hook of a video.
When Claude delivered a first edit, he gave very specific, timestamped feedback.
About 10 days in, he took one hook and focused on sound. He went through eight to 10 different sound effects before finding what worked, because Claude’s first version was too busy for his taste.
He gave direct feedback like “reduce the volume of that” or “I don’t like this sound at all, take it out.” Slowly, Claude learned what he liked.
Here is the key part. After you give feedback, tell Claude to save what it learned to the skill itself. That way, every new edit improves the system.
In the next session, Christian loved that Claude had pulled a specific logo and paired it with a sound effect, but he wanted the logo larger. He also changed the color palette for some animations.
The more videos he made, the better it got. Today his skill has 36 rules it has to follow.
He also keeps adding new animations to the library, and fine-tunes them each time he uses them.
Every note he gives is a mistake Claude doesn’t repeat. That is how he gets exactly the output he wants.
Claude Code Video Editing in Action
To show the system working, Christian added a raw hook he had recorded that day, mistakes and all, and typed a simple prompt.
He asked Claude to use his YouTube editing skill to edit the hook. He also told it to show a zoom, add a sound effect, and showcase three of the most impressive animations from the skill’s library.
Claude found the skill and ran it. It probed the clip, copied it, and transcribed it with Parakeet.
The process takes anywhere from 10 to 20 minutes. The result was a first version that cut 2 minutes and 46 seconds of raw footage down to 31 seconds.
It was nearly spot-on, with a few small changes to make. His first note was to add a suspense sound track throughout the hook.
He usually goes through the whole clip with timestamps, such as “at 22 seconds, you did this, I’d like it to be like that.” Feedback can cover the look, the sound, the overall feel, or the way it cuts.
At the end, he asks Claude to learn from what they just did and add it to the skill.
The hook at the start of his video was the final result of that one-shot run.
He used to edit in sections. Now he dumps in the entire footage. If he recorded in Tella, he just tells Claude the file names, since it has direct access.
After recording the video, he feeds in the raw footage and goes to sleep. The next morning, he watches the edit, gives a few notes, and gets the final version.
As he puts it, he is the director and Claude is the editor.
How Much Does Claude Code Video Editing Cost?
Here is the breakdown, down to the dollar:
| Tool | Cost |
|---|---|
| Claude (Max plan) | $200 per month flat. His last video used at most $24 of tokens. |
| Hyperframes | Free |
| Parakeet | Free |
| Tella (annual plan) | $13 per month, roughly $2 per video |
| Epidemic Sound (annual plan) | $9.99 per month, roughly $1 per video |
| Total | About $27 per video |
Christian’s $200 Max plan is a flat fee that he uses for all of his work, not only video editing. To get the per-video number, he had Claude measure the tokens used for editing.
For comparison, a good YouTube editor costs anywhere from $250 to $1,000 or more per video. That is hundreds of dollars saved on each one.
Why Automating Video Editing Saves More Than Money
The biggest benefit for Christian is speed. He can now plan, shoot, edit, and release a video on the same day.
Before, the process took between 3 and 7 days, depending on the editor.
Revisions went back and forth, and human error often meant repeating notes that were never applied from one video to the next.
With a skill that saves every note, those repeated mistakes disappear.
If you want to skip the trial and error, you can get his editing skill for free in the WeAreNoCode community.
Can Claude Code edit videos?
Yes. Claude Code can edit videos when it is connected to tools like Parakeet for transcription, Hyperframes for animations, Tella for screen recordings, Epidemic Sound for music and sound effects, and FFmpeg for final checks. Claude Code coordinates all of them from your computer.
Do I need to know how to code to automate video editing with Claude Code?
No. Christian Peverelli says he doesn’t know how to code or edit. Claude Code can set up the system itself and can even write the editing skill file for you.
How much does it cost to automate video editing with Claude Code?
In Christian’s setup, it costs about $27 per video. That is up to $24 in Claude usage, about $2 for Tella, and about $1 for Epidemic Sound. Hyperframes and Parakeet are free. A human YouTube editor typically costs $250 to $1,000 or more per video.
What is a Claude Code skill?
A skill is a file of instructions that Claude reads before it does a job. For video editing, it holds your style preferences, such as colors, fonts, animations, and sounds. Christian’s skill, YouTube-Edit, is about 1,400 lines long and has 36 rules.
What Claude plan do I need for Claude Code video editing?
You need a paid plan. Christian recommends starting with the $20 plan. He personally uses the $200 Max plan because he uses Claude for all of his work.
Advertisements
David
![You ask your AI a simple question. It gives you a confident answer with specific details. Then you fact-check it. Everything's wrong. Welcome to AI hallucinations. The problem that makes you second-guess every response, even when the AI sounds completely certain. GPT-5 and Claude 4 are the most advanced language models available. But they still make things up. The question isn't if they'll hallucinate. It's how often, why, and what you can do about it. I've spent three months testing both models. Running identical prompts. Fact-checking responses. Tracking hallucination rates across different tasks. Here's what I found, which model is more reliable, and how to reduce hallucinations in both. The Hallucination Reality Check First, let's be clear about what we're dealing with. AI hallucinations aren't bugs. They're features of how these models work. They predict the next word based on patterns in their training data. Sometimes those predictions sound right but are completely wrong. Both GPT-5 and Claude 4 hallucinate. But they do it differently, in different situations, at different rates. Understanding these differences helps you choose the right model and use it correctly. GPT-5 Hallucination Patterns: What I Found GPT-5 is fast, creative, and confident. Sometimes too confident. Where GPT-5 hallucinates most: Specific facts and figures: Ask for statistics, dates, or numbers and GPT-5 will give you precise answers. Often wrong. I asked it for the population of mid-sized cities. It gave me numbers that were off by 20-30%. But it presented them with complete certainty. Recent events: Anything after its knowledge cutoff is a gamble. GPT-5 will fill gaps with plausible-sounding information rather than admitting uncertainty. Asked about a tech company's Q3 2024 earnings. It gave me revenue figures that sounded reasonable. All fabricated. Citations and sources: Request sources and GPT-5 will provide them. Book titles. Article names. URLs. Many don't exist. I tested this with academic research requests. About 40% of the citations it generated were completely made up. Real-sounding titles. Fake papers. Technical specifications: Product specs, API details, version features. GPT-5 blends what it knows with what sounds right. Asked about specific features in React 19. It listed capabilities that don't exist yet, mixed with real features. Where GPT-5 is reliable: General knowledge: Common facts, well-documented history, widely-known information. Here, GPT-5 is solid. Code patterns: Standard programming solutions and common implementations. It's seen millions of examples. Creative work: When accuracy doesn't matter, hallucinations don't hurt. Writing fiction? GPT-5 is fine. Conceptual explanations: How things work in general. Principles and concepts rather than specific facts. Hallucination rate in my testing: Factual questions: 25-35% contained at least one hallucination Technical details: 20-25% hallucination rate Recent events: 40-50% hallucination rate General knowledge: 10-15% hallucination rate Claude 4 Hallucination Patterns: What I Found Claude 4 takes a different approach. It's more cautious, more likely to express uncertainty, and generally more accurate on facts. Where Claude 4 hallucinates most: Obscure information: When dealing with niche topics or rare details, Claude sometimes fills gaps rather than admitting it doesn't know. I asked about a small regional festival. Claude gave me dates and details that sounded specific. Couldn't verify any of it. Connecting unrelated facts: Claude is good at reasoning, but sometimes makes logical leaps that aren't supported. Asked about correlations in dataset. Claude confidently explained relationships that weren't actually there. Completing patterns: When you give it partial information, Claude tries to complete it. Sometimes those completions are invented. Started describing a hypothetical product. Claude added features and specifications I never mentioned, treating them as real. Where Claude 4 is reliable: Factual caution: Claude often says "I'm not certain" or "Based on my training data" rather than making things up. This is huge. Reasoning through problems: When Claude shows its thinking process (extended thinking), hallucinations drop significantly. Admitting limitations: Claude is more likely to say "I don't have information about that" than to fabricate an answer. Technical accuracy: For well-documented technical topics, Claude is consistently more accurate than GPT-5. Hallucination rate in my testing: Factual questions: 15-20% contained at least one hallucination Technical details: 10-15% hallucination rate Recent events: 25-30% hallucination rate General knowledge: 5-10% hallucination rate Claude 4 hallucinates roughly 40% less than GPT-5 across most categories. The Key Difference: Confidence vs. Caution The biggest difference isn't just accuracy. It's how each model handles uncertainty. GPT-5 approach: Always gives you an answer. Even when it's not sure. Confidence over accuracy. Claude 4 approach: More likely to express uncertainty or admit gaps. Accuracy over confidence. This matters in practice. GPT-5 feels more helpful because it never says "I don't know." But that helpfulness includes making things up. Claude 4 feels more honest because it admits limitations. But sometimes you want an answer, even if it's imperfect. Choose based on your use case. Need creativity and don't care about perfect accuracy? GPT-5 works. Need facts you can trust? Claude 4 is safer. How to Reduce GPT-5 Hallucinations Here are the techniques that actually work for GPT-5: 1. Request Sources and Citations Bad prompt: "What's the average salary for data scientists in 2024?" Better prompt: "What's the average salary for data scientists in 2024? Please cite your sources and note if you're uncertain about any figures." When you ask for sources, GPT-5 is more careful. It won't eliminate hallucinations, but it reduces them by about 30% in my testing. 2. Use Step-by-Step Reasoning Bad prompt: "Is this investment strategy sound?" Better prompt: "Analyze this investment strategy step by step. First, identify the key assumptions. Then, evaluate each assumption. Finally, give your assessment." Breaking down reasoning reduces logical leaps and makes hallucinations more obvious. 3. Set Conservative Parameters Use lower temperature settings (0.3-0.5) for factual tasks. Higher creativity means more hallucinations. Use this in the API or ask ChatGPT to "be conservative and fact-focused" in your prompt. 4. Verify Recent Information Add to prompts: "If this information is after your knowledge cutoff date, please say so explicitly." This forces GPT-5 to acknowledge when it's guessing about recent events. 5. Request Confidence Levels Add to prompts: "Rate your confidence in this answer from 1-10 and explain why." GPT-5 will often rate itself lower on information it's less certain about. Not perfect, but helpful. 6. Use Negative Examples Add to prompts: "Do not make up citations, dates, or statistics. If you're not certain about a specific detail, say so." Explicit instructions to avoid hallucinations help. Not completely, but measurably. 7. Double-Check Specific Claims Never trust specific numbers, dates, citations, or recent events without verification. Period. How to Reduce Claude 4 Hallucinations Claude 4 needs different techniques because it hallucinates differently: 1. Use Extended Thinking When available, use Claude's extended thinking mode for complex queries. The hallucination rate drops by 50% when Claude shows its reasoning. Standard prompt: "Explain this technical concept." Extended thinking prompt: "Take time to think through this technical concept carefully. Show your reasoning process." 2. Ask for Uncertainty Markers Add to prompts: "Please mark any statements you're uncertain about with [uncertain] tags." Claude is honest about uncertainty when you ask. This is its strength. 3. Request Reasoning Chains Better prompt: "Explain your reasoning for this conclusion. What evidence supports it? What evidence might contradict it?" Claude's hallucinations often happen in its conclusions, not its reasoning. Make it show both. 4. Avoid Leading Questions Claude sometimes tries to agree with your assumptions. Frame questions neutrally. Bad: "This data shows X is true, right?" Better: "What does this data actually show? Consider alternative interpretations." 5. Use Structured Outputs Add to prompts: "Format your response as: Facts (what you're certain about), Inferences (what you're reasoning toward), Uncertainties (what you don't know)." Structure reduces the chance of mixing facts with speculation. 6. Leverage Citations Mode When Claude cites sources (in modes where this is available), hallucinations drop significantly. Request citations whenever possible. 7. Challenge Confident Claims When Claude states something definitively, push back: "How certain are you about that? What would you need to verify it?" Claude will often back down from overconfident claims when challenged. Techniques That Work for Both Models Some strategies reduce hallucinations in both GPT-5 and Claude 4: 1. Provide Context The more context you give, the less the AI needs to guess. Bad: "What's the best framework?" Better: "I'm building a real-time dashboard with 10K concurrent users. Data updates every second. Team knows React. What framework should I use and why?" 2. Break Complex Questions Down Don't ask one massive question. Break it into steps. Instead of: "Design a complete system architecture for my app." Do: "First, what are the key components of this system? [Wait for response] Now, how should these components communicate? [Wait for response] What database makes sense given these requirements?" 3. Verify Critical Information For anything important, use multiple approaches: Ask the same question differently and compare answers Request the AI to fact-check itself Cross-reference with other sources Use web search features when available 4. Use Specific Constraints Add to prompts: "Only provide information you have high confidence in. For anything else, say 'I'm not certain about this' explicitly." Both models respond better to explicit guardrails. 5. Test with Known Answers Before trusting a model on unknown information, test it with questions you know the answer to. See how it handles uncertainty and accuracy. When Each Model is the Better Choice Choose GPT-5 when: Speed matters more than perfect accuracy You're brainstorming or being creative You can easily verify the output You need broad general knowledge You want conversational, confident responses Choose Claude 4 when: Accuracy is critical You're working with technical details You need transparent reasoning You value honesty about limitations You can wait for extended thinking on complex problems The Reality: Perfect Accuracy Doesn't Exist Here's the truth: You cannot eliminate AI hallucinations completely. Not with GPT-5. Not with Claude 4. Not with any language model. These tools predict text. They don't verify truth. That's not how they work. The best you can do is: Understand when each model is likely to hallucinate Use prompting techniques that reduce the rate Verify anything important Choose the right model for each task Set realistic expectations In my testing, good prompting techniques reduced hallucinations by 40-60%. That's significant. But it's not elimination. The winners aren't people who eliminate hallucinations. They're people who work with these limitations intelligently. The Bottom Line Claude 4 hallucinates less than GPT-5. About 40% less in my testing. It's more cautious, more honest about uncertainty, and more accurate on facts. But GPT-5 is faster, more confident, and sometimes that's what you need. Both require the same approach: Smart prompting. Healthy skepticism. Verification of critical facts. Use Claude 4 when accuracy matters. Use GPT-5 when speed and creativity matter. Use verification for both. The future of AI isn't hallucination-free responses. It's users who know how to work with imperfect tools to get reliable results. GPT-5 vs Claude 4: Which Hallucinates Less and How to Fix Both](https://ofemwire.com/wp-content/uploads/2026/01/pexels-photo-16245253-380x285.jpeg "GPT-5 vs Claude 4: Which Hallucinates Less and How to Fix Both 5")
![You ask your AI a simple question. It gives you a confident answer with specific details. Then you fact-check it. Everything's wrong. Welcome to AI hallucinations. The problem that makes you second-guess every response, even when the AI sounds completely certain. GPT-5 and Claude 4 are the most advanced language models available. But they still make things up. The question isn't if they'll hallucinate. It's how often, why, and what you can do about it. I've spent three months testing both models. Running identical prompts. Fact-checking responses. Tracking hallucination rates across different tasks. Here's what I found, which model is more reliable, and how to reduce hallucinations in both. The Hallucination Reality Check First, let's be clear about what we're dealing with. AI hallucinations aren't bugs. They're features of how these models work. They predict the next word based on patterns in their training data. Sometimes those predictions sound right but are completely wrong. Both GPT-5 and Claude 4 hallucinate. But they do it differently, in different situations, at different rates. Understanding these differences helps you choose the right model and use it correctly. GPT-5 Hallucination Patterns: What I Found GPT-5 is fast, creative, and confident. Sometimes too confident. Where GPT-5 hallucinates most: Specific facts and figures: Ask for statistics, dates, or numbers and GPT-5 will give you precise answers. Often wrong. I asked it for the population of mid-sized cities. It gave me numbers that were off by 20-30%. But it presented them with complete certainty. Recent events: Anything after its knowledge cutoff is a gamble. GPT-5 will fill gaps with plausible-sounding information rather than admitting uncertainty. Asked about a tech company's Q3 2024 earnings. It gave me revenue figures that sounded reasonable. All fabricated. Citations and sources: Request sources and GPT-5 will provide them. Book titles. Article names. URLs. Many don't exist. I tested this with academic research requests. About 40% of the citations it generated were completely made up. Real-sounding titles. Fake papers. Technical specifications: Product specs, API details, version features. GPT-5 blends what it knows with what sounds right. Asked about specific features in React 19. It listed capabilities that don't exist yet, mixed with real features. Where GPT-5 is reliable: General knowledge: Common facts, well-documented history, widely-known information. Here, GPT-5 is solid. Code patterns: Standard programming solutions and common implementations. It's seen millions of examples. Creative work: When accuracy doesn't matter, hallucinations don't hurt. Writing fiction? GPT-5 is fine. Conceptual explanations: How things work in general. Principles and concepts rather than specific facts. Hallucination rate in my testing: Factual questions: 25-35% contained at least one hallucination Technical details: 20-25% hallucination rate Recent events: 40-50% hallucination rate General knowledge: 10-15% hallucination rate Claude 4 Hallucination Patterns: What I Found Claude 4 takes a different approach. It's more cautious, more likely to express uncertainty, and generally more accurate on facts. Where Claude 4 hallucinates most: Obscure information: When dealing with niche topics or rare details, Claude sometimes fills gaps rather than admitting it doesn't know. I asked about a small regional festival. Claude gave me dates and details that sounded specific. Couldn't verify any of it. Connecting unrelated facts: Claude is good at reasoning, but sometimes makes logical leaps that aren't supported. Asked about correlations in dataset. Claude confidently explained relationships that weren't actually there. Completing patterns: When you give it partial information, Claude tries to complete it. Sometimes those completions are invented. Started describing a hypothetical product. Claude added features and specifications I never mentioned, treating them as real. Where Claude 4 is reliable: Factual caution: Claude often says "I'm not certain" or "Based on my training data" rather than making things up. This is huge. Reasoning through problems: When Claude shows its thinking process (extended thinking), hallucinations drop significantly. Admitting limitations: Claude is more likely to say "I don't have information about that" than to fabricate an answer. Technical accuracy: For well-documented technical topics, Claude is consistently more accurate than GPT-5. Hallucination rate in my testing: Factual questions: 15-20% contained at least one hallucination Technical details: 10-15% hallucination rate Recent events: 25-30% hallucination rate General knowledge: 5-10% hallucination rate Claude 4 hallucinates roughly 40% less than GPT-5 across most categories. The Key Difference: Confidence vs. Caution The biggest difference isn't just accuracy. It's how each model handles uncertainty. GPT-5 approach: Always gives you an answer. Even when it's not sure. Confidence over accuracy. Claude 4 approach: More likely to express uncertainty or admit gaps. Accuracy over confidence. This matters in practice. GPT-5 feels more helpful because it never says "I don't know." But that helpfulness includes making things up. Claude 4 feels more honest because it admits limitations. But sometimes you want an answer, even if it's imperfect. Choose based on your use case. Need creativity and don't care about perfect accuracy? GPT-5 works. Need facts you can trust? Claude 4 is safer. How to Reduce GPT-5 Hallucinations Here are the techniques that actually work for GPT-5: 1. Request Sources and Citations Bad prompt: "What's the average salary for data scientists in 2024?" Better prompt: "What's the average salary for data scientists in 2024? Please cite your sources and note if you're uncertain about any figures." When you ask for sources, GPT-5 is more careful. It won't eliminate hallucinations, but it reduces them by about 30% in my testing. 2. Use Step-by-Step Reasoning Bad prompt: "Is this investment strategy sound?" Better prompt: "Analyze this investment strategy step by step. First, identify the key assumptions. Then, evaluate each assumption. Finally, give your assessment." Breaking down reasoning reduces logical leaps and makes hallucinations more obvious. 3. Set Conservative Parameters Use lower temperature settings (0.3-0.5) for factual tasks. Higher creativity means more hallucinations. Use this in the API or ask ChatGPT to "be conservative and fact-focused" in your prompt. 4. Verify Recent Information Add to prompts: "If this information is after your knowledge cutoff date, please say so explicitly." This forces GPT-5 to acknowledge when it's guessing about recent events. 5. Request Confidence Levels Add to prompts: "Rate your confidence in this answer from 1-10 and explain why." GPT-5 will often rate itself lower on information it's less certain about. Not perfect, but helpful. 6. Use Negative Examples Add to prompts: "Do not make up citations, dates, or statistics. If you're not certain about a specific detail, say so." Explicit instructions to avoid hallucinations help. Not completely, but measurably. 7. Double-Check Specific Claims Never trust specific numbers, dates, citations, or recent events without verification. Period. How to Reduce Claude 4 Hallucinations Claude 4 needs different techniques because it hallucinates differently: 1. Use Extended Thinking When available, use Claude's extended thinking mode for complex queries. The hallucination rate drops by 50% when Claude shows its reasoning. Standard prompt: "Explain this technical concept." Extended thinking prompt: "Take time to think through this technical concept carefully. Show your reasoning process." 2. Ask for Uncertainty Markers Add to prompts: "Please mark any statements you're uncertain about with [uncertain] tags." Claude is honest about uncertainty when you ask. This is its strength. 3. Request Reasoning Chains Better prompt: "Explain your reasoning for this conclusion. What evidence supports it? What evidence might contradict it?" Claude's hallucinations often happen in its conclusions, not its reasoning. Make it show both. 4. Avoid Leading Questions Claude sometimes tries to agree with your assumptions. Frame questions neutrally. Bad: "This data shows X is true, right?" Better: "What does this data actually show? Consider alternative interpretations." 5. Use Structured Outputs Add to prompts: "Format your response as: Facts (what you're certain about), Inferences (what you're reasoning toward), Uncertainties (what you don't know)." Structure reduces the chance of mixing facts with speculation. 6. Leverage Citations Mode When Claude cites sources (in modes where this is available), hallucinations drop significantly. Request citations whenever possible. 7. Challenge Confident Claims When Claude states something definitively, push back: "How certain are you about that? What would you need to verify it?" Claude will often back down from overconfident claims when challenged. Techniques That Work for Both Models Some strategies reduce hallucinations in both GPT-5 and Claude 4: 1. Provide Context The more context you give, the less the AI needs to guess. Bad: "What's the best framework?" Better: "I'm building a real-time dashboard with 10K concurrent users. Data updates every second. Team knows React. What framework should I use and why?" 2. Break Complex Questions Down Don't ask one massive question. Break it into steps. Instead of: "Design a complete system architecture for my app." Do: "First, what are the key components of this system? [Wait for response] Now, how should these components communicate? [Wait for response] What database makes sense given these requirements?" 3. Verify Critical Information For anything important, use multiple approaches: Ask the same question differently and compare answers Request the AI to fact-check itself Cross-reference with other sources Use web search features when available 4. Use Specific Constraints Add to prompts: "Only provide information you have high confidence in. For anything else, say 'I'm not certain about this' explicitly." Both models respond better to explicit guardrails. 5. Test with Known Answers Before trusting a model on unknown information, test it with questions you know the answer to. See how it handles uncertainty and accuracy. When Each Model is the Better Choice Choose GPT-5 when: Speed matters more than perfect accuracy You're brainstorming or being creative You can easily verify the output You need broad general knowledge You want conversational, confident responses Choose Claude 4 when: Accuracy is critical You're working with technical details You need transparent reasoning You value honesty about limitations You can wait for extended thinking on complex problems The Reality: Perfect Accuracy Doesn't Exist Here's the truth: You cannot eliminate AI hallucinations completely. Not with GPT-5. Not with Claude 4. Not with any language model. These tools predict text. They don't verify truth. That's not how they work. The best you can do is: Understand when each model is likely to hallucinate Use prompting techniques that reduce the rate Verify anything important Choose the right model for each task Set realistic expectations In my testing, good prompting techniques reduced hallucinations by 40-60%. That's significant. But it's not elimination. The winners aren't people who eliminate hallucinations. They're people who work with these limitations intelligently. The Bottom Line Claude 4 hallucinates less than GPT-5. About 40% less in my testing. It's more cautious, more honest about uncertainty, and more accurate on facts. But GPT-5 is faster, more confident, and sometimes that's what you need. Both require the same approach: Smart prompting. Healthy skepticism. Verification of critical facts. Use Claude 4 when accuracy matters. Use GPT-5 when speed and creativity matter. Use verification for both. The future of AI isn't hallucination-free responses. It's users who know how to work with imperfect tools to get reliable results. GPT-5 vs Claude 4: Which Hallucinates Less and How to Fix Both](https://ofemwire.com/wp-content/uploads/2026/01/pexels-photo-16245253-300x225.jpeg "GPT-5 vs Claude 4: Which Hallucinates Less and How to Fix Both 11")