Skip to content
GrowthVector.io

Tips

Prompting Techniques

01

Chain-of-Thought

Force the model to show its work for complex reasoning

Make the model break down complex problems into explicit reasoning steps before answering. Dramatically improves accuracy on multi-step analysis, calculations, and strategic decisions.

How It Works
  • Add 'Let's think step by step' or 'Think through this step by step' to any prompt
  • Model generates intermediate reasoning tokens before the final answer
  • Each step provides context for the next, reducing logic errors
  • Works best on problems with clear cause-and-effect relationships

CoT was originally observed in 100B+ parameter models, but modern architectures and training techniques (like distillation and fine-tuning) now enable effective CoT reasoning in much smaller models. Use with GPT-5.5, Claude Opus 4.8 / Sonnet 5, Gemini 3+, or their smaller variants.

Best Use Cases
ROI and attribution calculationsCampaign budget allocation decisionsSEO keyword prioritizationMulti-channel strategy planningComplex A/B test analysis
Example Prompt Pattern
Our Facebook campaign data: - Ad spend: $5,000 - Conversions: 125 - Average order value: $85 - Product margin: 40% Should we scale this campaign? Think through this step by step: Step 1: Calculate total revenue Step 2: Calculate ROAS Step 3: Calculate profit after ad spend Step 4: Determine if sustainable to scale Step 5: Make recommendation
02

Generated Knowledge

Dump relevant facts before reasoning

Two-step process: first ask the model to generate relevant knowledge, then use that knowledge as context for the actual task. Reduces hallucination by making supporting facts explicit.

How It Works
  • Step 1: Ask model to generate 5-10 relevant facts about the topic
  • Step 2: Use those facts as context for your actual question
  • Separates knowledge retrieval from reasoning
  • Forces model to articulate what it knows before applying it

Grounding in explicit facts cuts hallucination. The model can't make up answers when it must first state what it knows.

Best Use Cases
Complex topics requiring domain knowledgeCompliance and policy questions (GDPR, legal)Technical reviews needing best practicesBrand persona developmentWhen responses feel generic or lack depth
Example Prompt Pattern
Step 1 - Generate Knowledge: "Generate 5 key facts about GDPR email marketing compliance" [Model generates facts] Step 2 - Apply Knowledge: "Using the facts above, evaluate if this campaign is compliant: - Purchased email list - No unsubscribe link - Auto-enrolled in newsletter"
03

Reflexion

Learn from failures with episodic memory

When the model fails or produces suboptimal output, prompt it to reflect on why it failed. Store reflections in memory. On next attempt, the model reviews past failures to avoid repeating mistakes.

How It Works
  • Model attempts task and either succeeds or fails
  • If failed: prompt 'Why did this fail? What would you do differently?'
  • Store reflection in a conversation memory buffer
  • On retry: include past reflections in the prompt
  • Model avoids repeating the same errors
Best Use Cases
Iterative content optimizationA/B test hypothesis refinementCampaign strategy adjustmentsDebugging complex workflowsAny task requiring multiple attempts to perfect
Workflow
  1. 1Attempt 1: Model tries task, fails
  2. 2Reflection: 'I failed because I didn't verify the data source'
  3. 3Attempt 2: Given reflection, model now checks data source first
  4. 4Result: Success through iterative self-correction
Example Prompt Pattern
Attempt 1: [Model output was incorrect] Reflect: "Analyze why your previous answer was wrong. What did you miss? What would you do differently?" [Model reflects] Attempt 2: "Now try again, keeping your reflection in mind: [Same task]"

Advanced: Maintain a running 'lessons learned' document. Feed it back into future prompts to build institutional knowledge.

04

Graph of Thoughts

Merge parallel reasoning paths into unified solutions

Like Tree of Thoughts but reasoning paths can merge, not just branch. Allows brainstorming multiple ideas in parallel, developing them independently, then synthesizing into a final solution.

How It Works
  • Branch: Generate multiple independent reasoning paths
  • Develop: Work on each path in parallel
  • Aggregate: Merge insights from different paths
  • Refine: Loop back to improve specific nodes
  • Synthesize: Combine into unified final output
Best Use Cases
Strategic content repurposing (whitepaper to multi-channel)Campaign theme developmentMulti-stakeholder messagingComplex synthesis tasksWhen you need diverse perspectives merged
Workflow
  1. 1Generate 3 marketing angles for product launch
  2. 2Develop each angle with supporting points
  3. 3Identify best elements from each angle
  4. 4Merge elements into one cohesive campaign message
  5. 5Refine the merged message for clarity
Example Prompt Pattern
Task: Turn whitepaper into month of content Node A: Extract data points Node B: Extract quotes Node C: Identify narrative themes Node D (merge A+C): Create infographics from data + themes Node E (merge B+C): Turn quotes into LinkedIn hooks + themes Node F (final): Combine D+E into content calendar

Limitation: Complex to manage manually. Best for high-value synthesis tasks, not routine work.

05

Few-Shot Prompting

Show 2-5 examples to enforce format and style

Provide 2-5 input-output examples before your actual task. The model pattern-matches to replicate the exact format, tone, and structure you want.

How It Works
  • Include 2-5 examples demonstrating the desired transformation
  • Model's attention mechanism copies the pattern from examples
  • Order matters: sometimes best examples last (recency bias)
  • Each example should match the complexity of your real task
Best Use Cases
Consistent ad copy formattingEmail tone transformationClassification tasks (intent, sentiment)UTM parameter naming standardsLanding page headline testing variants
Key Components
  1. 1Keep all examples in the same format
  2. 2Use diverse examples to show edge cases
  3. 3Match example tone to desired output tone
  4. 4Quality over quantity: 3 good examples beat 10 mediocre ones
Example Prompt Pattern
Classify customer inquiry type: Inquiry: "I ordered a blue shirt but received red" Type: Product Issue Inquiry: "What are your store hours on weekends?" Type: General Information Inquiry: "Can I get refund if I return within 30 days?" Type: Policy Question Inquiry: "Do you ship to Canada?" Type:

Limitation: Adds token cost (examples in every request). For high-volume use, consider fine-tuning instead.

06

ReAct (Reason + Act)

Connect the model to tools and real-time data

Give the model access to tools (search, calculator, APIs). It decides when to use them, creating a loop: Thought, Action, Observation, repeat.

How It Works
  • Define available tools with clear descriptions
  • Model generates a 'Thought' (what it needs to know)
  • Model emits an 'Action' (tool name + parameters)
  • System executes tool and returns 'Observation'
  • Model uses observation in next thought, creating a reasoning loop

ReAct is a foundation of many tool-using AI agents (though many modern agents use native function-calling rather than literal ReAct). It significantly reduces hallucination by grounding responses in external data, though it does not eliminate hallucination entirely.

Best Use Cases
SEO competitor research with live dataReal-time bid adjustments based on weather/eventsMarket share calculations from databasesPrice monitoring and dynamic pricingCampaign performance attribution
Workflow
  1. 1Model reasons: 'I need current SERP data for this keyword'
  2. 2Action: SEMrush_API.get_keyword_difficulty('enterprise AI')
  3. 3Observation: 'KD: 85 (Very Hard). Related: AI agents KD: 45'
  4. 4Model adjusts strategy based on real data
Example Prompt Pattern
Available tools: - SEARCH: Look up current information - CALCULATOR: Precise math calculations - SEMRUSH_API: Keyword and competitor data Task: Recommend 3 keywords for our project management SaaS campaign. Thought: I need current keyword difficulty scores Action: SEMRUSH_API.get_keywords("project management software") Observation: [API returns data] Thought: Now I'll analyze which have best difficulty/volume ratio
07

Prompt Chaining

Break complex tasks into focused single-step prompts

Split monolithic tasks into a sequence of simple prompts where each output feeds into the next. Each step has one job, making the chain reliable and debuggable.

How It Works
  • Design each prompt to do ONE thing well
  • Output of prompt N becomes input for prompt N+1
  • Add validation steps between chains when needed
  • Can branch or loop based on intermediate results
Best Use Cases
Content creation (research, outline, draft, optimize, promote)PPC campaign builds (keywords, ad groups, copy, landing pages)Email sequence development (triggers, sequence, copy, tests)Complex data processing pipelinesMulti-stage customer journey analysis
Key Components
  1. 1Keep each step simple and focused
  2. 2Define clean handoffs between steps
  3. 3Include checkpoints for quality control
  4. 4Document the flow for team reuse
Example Prompt Pattern
Chain 1: "List top 5 HR software questions from search data" then Chain 2: "Create blog outline addressing: [Chain 1 output]" then Chain 3: "Write full 2000-word post from outline: [Chain 2 output]" then Chain 4: "Optimize for SEO, add internal links: [Chain 3 output]" then Chain 5: "Create 5 LinkedIn posts promoting: [Chain 4 output]"

Advanced: Use chaining to insert human review steps. After Chain 2 (outline), get approval before Chain 3 (expensive full draft).

08

Tree of Thoughts

Explore multiple strategies, evaluate, and backtrack

Structure problem-solving as a search tree. Generate multiple options at each step, evaluate them, keep the best, and backtrack when paths fail. Like strategic planning or chess.

How It Works
  • Break challenge into decision points (nodes in a tree)
  • At each node, generate 3-5 possible approaches (branches)
  • Evaluate each branch's viability before proceeding
  • Pursue the most promising path, backtrack if it fails
  • Enables lookahead: simulate consequences before committing
Best Use Cases
Launch strategy development (channel mix decisions)Landing page optimization paths (what to test first)Crisis response planning (PR scenario modeling)Audience expansion strategies (who to target next)Content calendar themes (quarterly planning)
Workflow
  1. 1Define the strategic decision or creative challenge
  2. 2Branch: Generate multiple approaches at first decision point
  3. 3Evaluate: Score each branch (pros, cons, risks)
  4. 4Prune: Discard low-scoring branches
  5. 5Recurse: Continue branching on best path until solution emerges
Example Prompt Pattern
Challenge: Launch new feature with $50K in 60 days Level 1 - Channel Strategy: ├─ Branch A: Paid-First (Heavy PPC + retargeting) │ Pros: Fast, scalable │ Cons: Expensive, stops when budget stops │ ├─ Branch B: Content-Led (SEO + organic social) │ Pros: Sustainable, builds authority │ Cons: Slow, uncertain timing │ └─ Branch C: Hybrid (60% paid, 40% content) Pros: Balance speed and sustainability Cons: Split focus Evaluation: Given 60-day timeline, Branch A or C most viable. Level 2 - Expand Branch C: [Continue branching on selected path]

Limitation: Token-intensive: generates multiple paths. Use for high-stakes decisions only, not routine tasks.

09

Self-Consistency Checks

Slash hallucinations by reconciling multiple answers

Prompt the model for multiple answers, then have it compare and reconcile them to slash hallucinations and boost accuracy.

How It Works
  • Generate 3-5 responses to the same prompt using higher temperature (0.7-1.0) for diversity
  • Use majority voting: the most common answer wins for quantifiable outputs
  • For open-ended tasks, use Universal Self-Consistency (USC): ask the LLM to judge its own multiple outputs and select the best one
  • Combine with Chain-of-Thought prompting to expose different reasoning pathways
Example Prompt Pattern
Answer this question 5 times with different reasoning approaches: [Question] Then compare your answers and provide the most consistent conclusion.
10

Program-of-Thought

Numerical precision through code execution

Tell the model to solve problems with code or math tools; numerical precision skyrockets when it writes and runs functions.

How It Works
  • Instead of having the LLM compute in natural language, instruct it to generate executable code (typically Python)
  • The code is then run by an external interpreter for precise calculations
  • Eliminates 'in-head' computation errors that plague Chain-of-Thought for math
Best Use Cases
Complex math word problemsFinancial calculations (FinQA-style)Any task requiring numerical precision with large numbersStatistical analysis and data manipulation
Example Prompt Pattern
Solve this problem by writing Python code. Show your reasoning in comments, then provide the executable code: [Math Problem] Use SymPy for symbolic math if needed.

Limitation: Only effective for quantitative/structured problems that can be represented as code.

11

Plan-and-Solve Workflow

Step-by-step planning for better results

Make the model outline a step-by-step plan before executing any task for clearer logic and better results.

How It Works
  • Phase 1: Have the model create a detailed plan with numbered steps
  • Phase 2: Execute each step sequentially, checking results
  • Phase 3: Evaluate and refine iteratively
Key Components
  1. 1Define goal explicitly with scope and desired outcome
  2. 2Provide all relevant context and constraints upfront
  3. 3Specify output format requirements
  4. 4Use few-shot examples when possible
  5. 5Iterate: test, evaluate, refine, repeat
Example Prompt Pattern
Before completing this task, create a detailed plan: 1. List all steps needed to complete [task] 2. Identify potential challenges at each step 3. Only after I approve the plan, execute step by step Task: [Description]
12

Structural Guardrails

A large portion of Claude's system prompt is guardrails for a reason

Wrap your instructions with explicit edge cases, fallbacks, and output formats. A large portion of Claude's system prompt is dedicated to guardrails for a reason.

Types of Guardrails
Input validation:"If the user provides [X], respond with [Y]"
Output formatting:"Always respond in JSON with these fields: [...]"
Scope boundaries:"Only answer questions about [topic]. For anything else, say [fallback message]"
Safety constraints:"Never generate content that [list prohibited outputs]"
Error handling:"If you cannot complete the task, explain why and suggest alternatives"
Example Prompt Pattern
<system> You are [role]. Your purpose is [goal]. RULES: - Always [required behavior] - Never [prohibited behavior] - If [edge case], then [specific response] OUTPUT FORMAT: [Exact structure required] FALLBACK: If you cannot fulfill a request, respond: "[message]" </system>
13

Context Positioning & Negatives

Beginning and end matter most

Put mission-critical instructions near the beginning and end of the prompt and include 'don't do this' examples to avoid failure modes.

Everything is about providing the LLM guardrails. It's almost better for it to know what NOT to do than what to do.

The Primacy-Recency Effect
  • LLMs pay most attention to the beginning and end of prompts
  • Place your most important instructions near the start (sets the frame)
  • Repeat critical constraints near the end (recency bias)
  • Middle sections can contain context, examples, and details
Using Negative Examples
  • Show the model what BAD output looks like, then good output
  • 'Don't do this' examples prevent common failure modes
  • Format: ❌ WRONG: [bad example] followed by ✅ RIGHT: [good example]
Example Prompt Pattern
[CRITICAL INSTRUCTION - What you must do] [Context and background] [Examples - both positive and negative] [Details and specifications] [REMINDER - Repeat the critical instruction] Remember: [Most important constraint]
14

Meta-Prompting

The AI knows its capabilities better than you

Ask the model to critique, clarify, and suggest improvements to your prompt; it knows its own capabilities better than you do.

How It Works
  • Before using a prompt for its intended task, ask the LLM to analyze it
  • The model can identify ambiguities, missing context, and potential misinterpretations
  • Let the AI rewrite your prompt based on its self-knowledge
Workflow
  1. 1Write your initial prompt
  2. 2Ask: 'Critique this prompt. What's unclear? What might cause you to fail?'
  3. 3Ask: 'Rewrite this prompt to be more effective for your capabilities'
  4. 4Compare and iterate
Example Prompt Pattern
I'm going to give you a prompt I want to use. Before executing it, I need you to: 1. Identify any ambiguous instructions 2. List what context is missing 3. Suggest 3 improvements to make it more effective 4. Rewrite the prompt incorporating your suggestions Original prompt: [Your prompt here]
15

Reverse-Prompting

Learn from the AI's perspective

Ask the model to generate a prompt that achieves a specific outcome, then analyze the prompt for insights into its capabilities and limitations.

How It Works
  • Show the model the desired OUTPUT and ask it to generate the INPUT (prompt) that would produce it
  • Reveals how the model 'thinks' about prompt-output relationships
  • Uncovers hidden capabilities and optimal phrasing
Best Use Cases
Learning optimal prompt structures for specific tasksDiscovering command patterns you wouldn't have thought ofCreating reusable prompt templates from successful outputsUnderstanding model limitations by seeing what prompts it considers necessary
Example Prompt Pattern
Here is the output I want to achieve: [Desired output example] Generate the prompt that would produce this output. Explain why you structured the prompt this way and what elements are critical for success.

Advanced: Chain reverse-prompting with meta-prompting: 1) Reverse-prompt to generate a prompt, 2) Meta-prompt to critique it, 3) Iterate to find the optimal structure.