SMART Framework for AI Prompts: A Complete Guide

SMART Framework for AI Prompts: The Complete Guide

You already know SMART goals. If you have ever worked in project management, marketing, or any corporate role, you have written objectives that are Specific, Measurable, Achievable, Relevant, and Time-bound. The same framework that makes goals effective also makes AI prompts effective. SMART framework AI prompts take the goal-setting methodology George T. Doran introduced in 1981 and apply it directly to how you communicate with ChatGPT, Claude, Gemini, and other language models.

The concept is straightforward: vague prompts produce vague outputs. SMART prompts produce focused, evaluable, actionable outputs. This guide walks through each component, shows you how to build a SMART prompt step by step, and gives you 6 real-world examples you can copy and adapt immediately. For a quick-reference version of the framework itself, see the SMART framework page.

What Is the SMART Framework for AI Prompts?

SMART is an acronym where each letter represents one quality your prompt should have:

George T. Doran first published the SMART acronym in the November 1981 issue of Management Review, in an article titled "There's a S.M.A.R.T. Way to Write Management's Goals and Objectives." The framework built on Peter Drucker's Management by Objectives (MBO) concept from 1954, and later drew support from Edwin Locke's Goal-Setting Theory research in the 1960s, which proved that specific, challenging goals drive higher performance.

The adaptation for AI prompting is natural. Every quality that makes a goal effective, clarity, measurability, feasibility, relevance, and temporal grounding, also makes a prompt effective. The difference is that instead of managing a team toward an objective, you are managing an AI toward a useful output.

Why SMART Works for AI Prompts

Most prompts fail because they are missing at least two of the five SMART qualities. Consider this prompt:

"Write me a marketing plan."

This tells the AI almost nothing. What kind of marketing? For what product? What audience? What budget? What timeframe? The AI fills in every blank with generic assumptions, and you get a generic response.

Now apply SMART thinking:

"Create a 90-day content marketing plan (Time-bound) for a B2B SaaS startup selling project management software (Specific) to mid-market engineering teams (Relevant). Include 12 blog topics with target keywords and estimated monthly search volume (Measurable). Focus on strategies a 2-person marketing team can execute with a $3,000 monthly budget (Achievable)."

The second prompt constrains the AI in all the right ways. Every SMART component eliminates a category of assumptions the AI would otherwise make on its own.

• Specific prevents the AI from guessing what you want
• Measurable gives you a way to judge whether the output is complete and useful
• Achievable keeps you from asking for things the AI cannot do well (real-time data, future predictions, physical actions)
• Relevant ensures the output serves your actual purpose, not a generic version of it
• Time-bound shapes scope, urgency, and the level of detail in the response

Step-by-Step: Building a SMART Prompt

Let me walk through building a SMART prompt from scratch. The task: you need help creating a hiring plan for your startup.

Step 1: Start With Specific

Before: "Help me with hiring."

After adding Specific: "Create a hiring plan for 3 software engineering roles (1 senior backend, 1 mid-level frontend, 1 DevOps engineer) at a Series A startup with 15 employees."

You have defined the exact deliverable, the roles, and the company context. The AI no longer has to guess.

Step 2: Add Measurable

After adding Measurable: "...Include a sourcing channel comparison with estimated cost-per-hire and time-to-fill for each channel. Provide a scoring rubric with 5 criteria for evaluating candidates at each interview stage."

Now you have criteria: cost-per-hire, time-to-fill, a scoring rubric with a specific number of criteria. You can check the output against these benchmarks.

Step 3: Confirm Achievable

After adding Achievable: "...Base recommendations on standard startup hiring practices. The company has one in-house recruiter and a $15,000 quarterly recruiting budget. Do not assume access to enterprise recruiting tools or external agencies."

This grounds the plan in reality. The AI will not suggest hiring a team of recruiters or using expensive platforms the startup cannot afford.

Step 4: Ensure Relevant

After adding Relevant: "...The engineering team uses Python, React, and AWS. Prioritize candidates who can contribute to product development within 2 weeks of onboarding. The company culture values autonomy and async communication, so filter for remote-friendly candidates."

Every detail here connects to what actually matters for this specific startup. The AI will tailor its advice to the tech stack, culture, and onboarding expectations.

Step 5: Set Time-bound

After adding Time-bound: "...The first hire (senior backend) should start by June 1, 2026. The remaining two roles should be filled by August 31, 2026. Provide a week-by-week recruitment timeline starting from April 1, 2026."

The complete prompt is now roughly 150 words. Each of those words carries information that eliminates ambiguity and drives a better response.

6 Real-World SMART Framework Examples

Here are 6 complete SMART prompts across different business domains. Each is tested and ready to customize. For more framework options, explore our guides on advanced prompt engineering techniques and the best AI prompt frameworks in 2026.

Example 1: Business Planning

Specific: Create a go-to-market strategy for launching a B2B invoicing
SaaS product targeting freelancers and small agencies with 1-10 employees.

Measurable: Include 5 customer acquisition channels ranked by estimated
CAC, a pricing comparison table with 3 tiers, and monthly user growth
projections for the first 12 months with specific numbers.

Achievable: The founding team is 2 people (1 developer, 1 marketer)
with $20,000 in pre-launch budget. Do not assume paid advertising
budgets above $2,000/month or enterprise sales capabilities.

Relevant: The target audience currently uses spreadsheets or free tools
like Wave. They are price-sensitive but will pay for time savings.
Focus on product-led growth and content marketing over outbound sales.

Time-bound: The product launches on July 1, 2026. Provide a pre-launch
timeline (April through June) and a post-launch plan through December
2026 with monthly milestones.

Example 2: Marketing Campaign

Specific: Design an email marketing campaign to re-engage 2,500
inactive subscribers who have not opened an email in 90+ days for an
online education platform selling professional development courses.

Measurable: Create a 4-email sequence. For each email, provide subject
line (under 50 characters), preview text (under 90 characters), body
copy (under 200 words), and one CTA. Target a 15% reactivation rate
and include A/B test variations for subject lines.

Achievable: The platform uses Mailchimp for email. Segmentation is
based on last open date and course category interest. No dynamic
personalization beyond first name and last course viewed.

Relevant: The audience is working professionals aged 28-45 who
purchased at least one course previously. Tone should be helpful and
direct, not salesy. Emphasize new course releases and skill relevance
rather than discounts.

Time-bound: The sequence should deploy over 14 days, with emails
spaced 3-4 days apart. Reference Q2 2026 course launches as the
primary hook.

Example 3: Project Kickoff

Specific: Write a project brief for redesigning the checkout flow of a
Shopify-based e-commerce store that sells custom furniture. The
redesign should reduce cart abandonment and support a new installment
payment option.

Measurable: Include 4 success metrics with current baselines and
target improvements (e.g., cart abandonment rate from 72% to under
60%). Provide a task breakdown with estimated hours for each
deliverable. List 5 specific UX improvements with expected impact.

Achievable: The team consists of 1 UX designer, 1 Shopify developer,
and 1 QA tester. Budget is 120 hours total. The store uses Shopify
Plus with standard Liquid templates, no headless setup.

Relevant: Average order value is $2,400. Customers often abandon
because they want financing options but the current checkout only
supports full payment. Mobile traffic is 65% of total but mobile
conversion is 40% lower than desktop.

Time-bound: Design phase runs April 1-15, 2026. Development runs
April 16 through May 15, 2026. QA and launch by May 31, 2026. Include
a week-by-week timeline.

Example 4: Hiring Plan

Specific: Create a structured interview process for hiring a Head of
Customer Success at a B2B SaaS company with 200 customers and $3M ARR.
Include job requirements, interview stages, and evaluation criteria.

Measurable: Define 6 core competencies with behavioral interview
questions for each. Provide a candidate scorecard with a 1-5 rating
scale per competency. Include 3 case study exercises with evaluation
rubrics.

Achievable: The hiring committee is the CEO, VP of Sales, and one
senior Customer Success Manager. The process should fit within 3
interview rounds (screen, deep-dive, final) to avoid candidate
drop-off.

Relevant: The company sells to mid-market B2B buyers with an average
contract value of $15K/year. Churn is currently 8% monthly, and the
primary reason customers leave is poor onboarding. The ideal candidate
should have experience reducing churn through structured onboarding
programs.

Time-bound: Job posting goes live April 7, 2026. Target an offer
extended by May 16, 2026. The new hire should start by June 2, 2026.
Provide a timeline for each interview round.

Example 5: Content Calendar

Specific: Build a 30-day social media content calendar for a DTC
skincare brand launching on Instagram, TikTok, and LinkedIn. The brand
targets women aged 25-40 who prioritize clean ingredients.

Measurable: Include 30 posts (one per day), each with platform,
content type (reel, carousel, story, static), caption (under 150
words), 5 hashtags, and suggested posting time. Target 20% engagement
rate increase over the current 2.3% baseline.

Achievable: The content team is 1 social media manager and 1 freelance
graphic designer. Video content budget is $500/month. All product
photography is already available. No influencer partnerships for this
cycle.

Relevant: The brand differentiates on transparency (full ingredient
lists, no greenwashing). Top competitors are Glossier and Drunk
Elephant. The audience responds best to educational content about
ingredients and before/after results.

Time-bound: The calendar covers April 1-30, 2026. A product launch
(new vitamin C serum) happens April 15, so the first two weeks should
build anticipation and the last two should drive purchases.

Example 6: Financial Analysis

Specific: Perform a unit economics analysis for a meal kit delivery
startup and identify the 3 most critical metrics that need improvement
before the company can raise a Series A round.

Measurable: Calculate and present: customer lifetime value (LTV),
customer acquisition cost (CAC), LTV:CAC ratio, gross margin per box,
contribution margin, and payback period. Benchmark each metric against
industry medians with a red/yellow/green health indicator. Provide
dollar-impact estimates for improving each critical metric by 10%.

Achievable: Use the following data: average order value $65, orders per
customer per month 3.2, average customer lifespan 5.8 months, monthly
churn 17%, blended CAC $89, COGS per box $38, fulfillment cost per box
$12, monthly marketing spend $45,000.

Relevant: The analysis is for a pitch deck appendix. The audience is
seed-stage investors who will compare these unit economics against
DoorDash, HelloFresh, and Factor75 at similar stages. Highlight
metrics where the company outperforms and flag where it underperforms.

Time-bound: The fundraise is planned for Q3 2026. The analysis should
reference current metrics (March 2026 baseline) and project where each
metric will be in 6 months if current trends continue versus if the
recommended improvements are made.

SMART vs TAG vs APE: Which Framework?

Choosing the right framework depends on what your prompt needs most. Here is a comparison to help you decide. For a deeper look at all available options, see our complete framework comparison for 2026.

• Your task has a timeline or deadline
• You need measurable success criteria
• The request requires balancing multiple constraints
• You want to evaluate the output against specific standards

• The task is straightforward and goal-oriented
• You do not need time context or measurability
• Speed matters more than precision
• A sentence or two is enough to define the task

• You need quick content generation
• The action and expected output are self-explanatory
• Feasibility and time context are irrelevant
• You want the simplest possible structure

• You need step-by-step instructions in the prompt
• The task requires a defined process with narrowing constraints
• Role assignment and end-goal clarity are both essential
• See the full RISEN tutorial with 7 examples

• The task requires analytical reasoning or step-by-step logic
• You are debugging code, solving math problems, or evaluating complex decisions
• See the chain-of-thought prompting guide

5 Common SMART Prompting Mistakes

1. Writing Specific Without Being Specific

The irony is real. People write "Specific: Create a marketing plan" and call it done. The Specific component needs actual specifics: what kind of marketing, what product, what audience, what format, what scope. If your Specific section could apply to a thousand different companies, it is not specific enough.

2. Using "Detailed" as a Measurable Criterion

"Make it detailed" is not measurable. "Include 5 recommendations, each with a cost estimate and implementation timeline" is measurable. Replace every subjective adjective (detailed, thorough, comprehensive, good) with a number, a format requirement, or a quality benchmark.

3. Asking the AI to Do Things It Cannot Do

Common Achievable failures: asking for real-time data, future predictions with certainty, access to private databases, or physical actions. The AI works with the knowledge it has and the context you provide. If your prompt requires information the AI does not have, provide it directly or adjust your expectations.

4. Forgetting the Audience in Relevant

A financial analysis for a CEO and the same analysis for a junior analyst should look completely different. Relevant is not just about topic alignment; it is about audience alignment. Always specify who will read the output and what they will do with it.

5. Treating Time-bound as Optional

Even prompts about "timeless" topics benefit from temporal context. "Analyze email marketing best practices" could mean practices from 2015 or 2026. "Analyze email marketing best practices as of Q1 2026, accounting for recent changes in Apple Mail privacy and Gmail inbox tabs" gives the AI a clear temporal frame that shapes every recommendation.

Tips for Different AI Models

SMART works across all major language models, but each has slight tendencies worth knowing.

ChatGPT (GPT-4, GPT-4o)

GPT-4 follows the Measurable component particularly well. When you specify exact numbers, format requirements, and quality benchmarks, GPT-4 tends to hit them precisely. It also responds well to the Time-bound component when asked for timelines and schedules. For more ChatGPT-specific techniques, see our ChatGPT prompting guide.

Claude (Claude 3.5, Claude 4)

Claude excels with the Relevant component. It is strong at understanding audience context and adjusting tone, depth, and vocabulary accordingly. Claude also tends to follow the Achievable constraints closely, rarely hallucinating capabilities it does not have. Pair SMART with explicit audience descriptions for best results with Claude.

Gemini

Gemini performs well with the Specific component, especially when you include structured format requirements (tables, numbered lists, comparison matrices). For Time-bound prompts that reference recent events, Gemini's access to current information can add value. Be extra clear with Measurable criteria, as Gemini sometimes defaults to longer outputs than requested.

General Tips Across Models

• Label each SMART component explicitly for complex prompts (500+ words)
• For shorter prompts, weave the five qualities into natural language
• Start with Specific and Time-bound, since these two components have the highest impact on output quality
• Use Measurable to prevent re-prompting: define what "done" looks like upfront

FAQ

Is the SMART framework only for business prompts?

No. SMART works for any prompt where clarity, measurability, and scope matter. Academic research requests, personal planning, creative project briefs, and technical documentation all benefit from the structure. The framework is domain-agnostic; it improves communication precision regardless of subject matter.

How is SMART different from just writing a longer prompt?

Length does not equal quality. A 500-word prompt can still be vague if it lacks measurable criteria or temporal context. SMART ensures that every element of your prompt serves a purpose. A 100-word SMART prompt will outperform a 500-word unstructured prompt because it covers five distinct dimensions of clarity rather than repeating the same instruction in different ways.

Can I combine SMART with other frameworks like RACE or RISEN?

Yes, and many experienced prompt engineers do exactly this. You can use RACE's Role component to set a professional persona, then apply SMART criteria to structure the rest of the prompt. The RISEN framework also pairs well with SMART, where RISEN provides the process structure and SMART provides the goal-setting discipline. Combining frameworks works best for complex, high-stakes prompts.

Do I need to label each component (S, M, A, R, T) in my prompt?

No. Labels help when the prompt is long or complex, but they are not required. SMART is a mental checklist, not a rigid template. A natural paragraph that covers all five qualities works just as well as labeled sections. The point is ensuring your prompt has all five qualities, not that it displays them with headers. For prompts under 100 words, weaving the components into natural language usually reads better.