13 ready-made prompts
AI Errors
Corrective prompts for the ways AI goes wrong: hallucination, fake citations, refusals, forgotten instructions and shallow answers.
Language models do not behave as if they know nothing — they write the most probable next sentence. The prompts in this category do not change the model's character; they make the output auditable. They demand sources, force uncertainty to be labelled, open refusals and pin down instructions that drift in long threads.
Prompt That Stops AI Hallucinations (Claim Labelling)
Splits the answer into three layers: verifiable fact, inference and guess. The model learns to say "I don't know", and you see which sentences you can trust.
Intermediate ChatGPTClaudeChatGPT Refuses to Answer? A Prompt That Opens the Refusal
Separates the cause of a refusal (safety filter, over-caution, vague request) and builds the right frame for each. It gets your work done without wandering.
Intermediate ChatGPTClaudePrompt for When the Model Forgets Your Instructions Mid-Chat
The model writes your rules back in its own words, then checks every reply against that list. Still working 40 messages later.
Intermediate ChatGPTClaudePrompt That Turns a Generic AI Answer Into Real Analysis
Bans empty lines like "it is important" and demands a claim, a counter-claim and the missing side for each point. Real analysis in three rounds.
Intermediate ChatGPTClaudePrompt to Fix AI Code: Catch Hallucinated APIs
Forces an inventory of every API the code uses, flags invented symbols and wrong signatures, and gives a verification plan before you run anything.
Advanced ChatGPTClaudeExcel Formula Not Working? Fix #REF!, #VALUE! and #NAME?
Breaks the formula into parts, finds separator, locale and range problems, and explains what each part returns. Systematic diagnosis for #REF!, #VALUE! and blanks.
Beginner ChatGPTClaudePrompt to Catch Outdated AI Answers (Knowledge Cutoff)
Separates fast-decaying facts (prices, versions, laws, personnel) and states where and when to verify each one.
Beginner ChatGPTClaudeInterview Answers That Sound AI-Written: Build Your Own
Connects the generic AI answer to your concrete experience, adds measurable results and pre-empts the three follow-up questions an interviewer will ask.
Beginner ChatGPTGeminiAI Recommending Products That Don't Exist: Verify the Model Number
Checks model number, release year and current availability separately, flags invented products and recommends a real alternative category.
Beginner ChatGPTClaudeAI-Written Emails That Sound Fake: Fix the Tone
Defines the relationship level, the single purpose and the response you want, then removes over-politeness, filler and stiff formulaic closings.
Beginner ChatGPTClaudePrompt That Fixes Broken JSON Output
Locks the schema and forbids anything outside the code block. Catches the trailing comma, the unquoted key and the unescaped quote before you spend an hour debugging them.
Advanced ChatGPTClaudePrompt That Stops the Repetition Loop
Maps every repeat into three types — same sentence, same idea, same sentence structure — then locates where the loop began. The rewrite comes with two word counts so you can see it actually worked.
Intermediate ChatGPTClaudePrompt That Stops Language Mixing
Stops the model answering in English halfway through. It separates real English sentences from technical terms that should stay in English, and reports how many words it moved in each bucket.
BeginnerFrequently asked
About AI Errors
Can hallucination be prevented completely?
No. The model's ability to notice it does not know something runs on the same mechanism as knowing it, so there is no 100% guarantee. What you can do is lower the risk: ask for sources, define an explicit "I don't know" option, and verify numbers and dates one by one in a second pass. These prompts do exactly that.
Why does the model refuse to answer?
Three different causes with different fixes: (1) a safety filter — the subject is legitimate but the wording triggered it, so the frame needs changing; (2) over-caution — the model is wary of a medical, legal or financial topic, so give it a role and boundaries; (3) a vague request — it cannot tell what you want, so impose the output format. The prompt asks which of the three you are testing.
Why does the model forget instructions in a long chat?
Every reply re-reads the whole conversation, and when the context window fills the earliest parts get summarised or truncated. Rules placed at the top disappear. The fix is not repeating them at the end of every message but making the model restate them in its own words and check each reply against that list.
Other categories