Jeff Bullas

Short answer: yes. AI can switch your writing into warm, witty, or authoritative tones on demand. The trick is giving it a clear brief and a reusable template so you stay in control of the voice.

Think of tone as a set of dials—warmth, wit, authority, energy, and formality. AI is very good at turning those dials if you tell it what you want, show a sample of your voice, and set a few boundaries.

What you need

• An AI writing tool (any mainstream one works)
• Two short samples of your writing (150–300 words each)
• One audience in mind (e.g., customers, partners, team)
• Ten minutes to set up your tone template

High-value shortcut: create a reusable “Tone Card” once, then apply it everywhere. It’s your personal style guide the AI can follow consistently.

Step-by-step

1. Build your Tone Card. Decide the tone dials and guardrails you want. Use the prompt below to create it.
2. Prime the AI with your samples. Paste 1–2 paragraphs of your own writing and ask it to extract your vocabulary, rhythm, and phrases. This makes the output feel like you, not a generic blog.
3. Use a “prompt sandwich.” Give Role, Reader, Goal, Constraints, and Examples before the Task. This prevents fluffy results.
4. Dial in tone by percentages. Ask for 70% authoritative, 20% warm, 10% witty—or any blend. Iterate until it reads right.
5. Polish. Add your signature phrases, cut filler, and ensure the reading level matches your audience.

Copy-paste Tone Card setup (use this once)

“You are my writing assistant. Study and store this Tone Card to use in all future responses unless I say otherwise.
Tone dials: Warmth=60, Wit=20, Authority=80, Energy=50, Formality=60, Empathy=70.
Voice rules: short paragraphs, plain English, no clichés, no exclamation points, reading level Grade 8–9. Prefer active voice, specific verbs, and concrete examples. Avoid jargon unless explained. Do not mention that you are an AI.
Do/Don’t: Do keep sentences under 22 words. Do include a clear takeaway. Don’t overpromise or use buzzwords. Don’t add emojis.
Signature phrases to allow: ‘Here’s the plan.’ ‘What to expect.’ ‘Quick win.’
Acknowledge this Tone Card with a one-line summary and ask me for a sample of my writing.”

Copy-paste prompt to analyze your voice (use with your samples)

“Analyze the following 2 short samples of my writing. List:
1) recurring phrases and word choices,
2) sentence rhythm (short/long),
3) formality level,
4) tone dials you infer (Warmth/Wit/Authority/Energy/Empathy out of 100).
Then create a 6-bullet ‘My Voice Notes’ I can reuse. Samples: [paste sample 1], [paste sample 2].”

Copy-paste prompt to rewrite in a chosen tone

“Role: Senior editor.
Reader: Busy professionals over 40.
Goal: Rewrite the text to be clear and engaging.
Tone blend: Authoritative 70, Warm 20, Witty 10.
Constraints: Grade 8–9 reading level; average sentence < 20 words; no clichés; no emojis; no self-references; 2 short paragraphs max; include one practical takeaway.
Task: Rewrite the following text and preserve all facts. Text: [paste text].”

Insider trick: the two-pass method

• Pass 1 (Clarity + Authority): Ask for a no-fluff, authoritative rewrite at low creativity. Keep it tight.
• Pass 2 (Warmth/Wit): Ask for a light layer of warmth or a single witty turn of phrase, with limits (e.g., “max one subtle metaphor”).

Example: one paragraph, three tones

Original: “Our newsletter includes updates and tips. It might help you be more productive.”

• Warm: “Think of this newsletter as a helpful nudge. Short tips, real examples, and one idea you can use today.”
• Witty: “Short reads, sharp tips. Less ‘someday’ advice, more ‘done by lunch.’”
• Authoritative: “Each issue delivers one proven tactic, a brief example, and a 5-minute action step. Designed for busy professionals.”

Mistakes to avoid (and quick fixes)

• Generic tone. Fix: Provide two short examples of copy you like and say “match the rhythm and concision.”
• Too clever. Fix: Cap wit at 10% and add “avoid puns and pop-culture references.”
• Jargon overload. Fix: “Explain terms in 7 words or fewer, or replace with plain English.”
• Overlong sentences. Fix: “Keep sentences under 20 words and paragraphs under 3 sentences.”
• Tone drift across sections. Fix: “Run a consistency check: list any sentences that break the Tone Card.”
• AI-speak. Fix: “No phrases like ‘leverage synergies,’ ‘in today’s world,’ or ‘as an AI.’”

What to expect

• First drafts will be 80% there. The last 20% is you adding specificity and trimming excess.
• Authority improves with evidence: ask for one stat, one example, or one named framework per piece.
• Wit works best as a spice, not the meal. One clever line is plenty.

15-minute action plan

1. Create your Tone Card using the first prompt (3 minutes).
2. Paste two samples, extract “My Voice Notes” (4 minutes).
3. Run a rewrite with the tone blend you need today (3 minutes).
4. Iterate once with a percentage tweak (2 minutes).
5. Save the Tone Card and Voice Notes as a preset for future work (3 minutes).

One more ready-to-use prompt: tone blend on demand

“Rewrite the text for [audience], with this blend: Warmth 50, Wit 15, Authority 70, Energy 60, Formality 50, Empathy 70. Constraints: Grade 8 reading level; keep original meaning; no clichés; include one clear action step; sentences under 20 words. Output: two versions. Text: [paste text].”

Use AI as your tone dial, not your ghostwriter. You bring the story and the judgment; the model brings speed and stylistic range. Set the dials once, then reuse them to sound warm, witty, or authoritative—whenever you need.

Smart question. You’re right to ask if AI can summarize competitor sites and pull out their positioning — it’s one of the fastest, lowest-risk wins you can get from AI.

Why this matters

• Websites hide positioning in plain sight: hero lines, pricing pages, case studies, and CTAs.
• AI can scan these quickly and standardize insights so you can compare apples to apples.
• Goal: a one-page battlecard per competitor plus a simple map of where you can win.

What you’ll need (15–45 minutes)

• 3–5 competitor URLs (homepage, pricing, features/solutions, and one case study).
• A browser and any AI chat that can read pasted text or browse pages.
• Optional: Reader Mode or “Print to PDF” to get clean text for pasting.

>

Do / Don’t checklist

• Do focus on: homepage hero, subheads, social proof, pricing/plan names, and the first 100 words of each page.
• Do grab About/Company language and any industry logos; these reveal target segments.
• Do standardize your output (same headings each time) so comparisons are clear.
• Do ask AI what’s missing (e.g., no pricing, weak proof, vague ROI).
• Don’t assume AI fetched every dynamic element; paste key text if a page blocks scraping.
• Don’t copy private or gated content; stick to public pages.
• Don’t stop at claims; ask for evidence sources (case studies, numbers) or mark as unsubstantiated.

Insider trick: Ask AI to infer positioning from subtle cues: plan names (Starter/Pro/Enterprise hint segments), hero image alt text, footer microcopy, awards badges, and repeated keywords in headlines. Also use search operators in your browser like: site:competitor.com pricing OR plans, site:competitor.com case study OR “customer story”.

Step-by-step: from URL to positioning map

1. Collect 3–5 key URLs per competitor: Home, Pricing, Features/Solutions, About, and one Case Study.
2. Capture text: Use Reader Mode or copy sections into your AI chat. If the tool can browse, give it the URLs and ask it to quote key snippets it’s using.
3. Standardize extraction: Run the prompt below for each competitor.
4. Compare: Feed all outputs to AI and ask for overlaps, gaps, and 2–3 “white space” angles you could own.
5. Draft your angle: Use the final prompt to create your own positioning and homepage hero ideas.

Copy-paste prompt (single competitor)

Analyze the website content below and extract their market positioning. Deliver a concise report in this exact outline and keep each bullet to one line:
1) Category and sub-category they want to own
2) Primary target segments (job titles, industries, company sizes)
3) Core pain points they focus on (3–5)
4) Value proposition and proof (claims + evidence cited)
5) Key features emphasized (not every feature; only proof-carrying ones)
6) Pricing and packaging signals (plan names, value levers)
7) Tone of voice and brand personality (2–3 adjectives)
8) Primary CTAs and offers
9) SEO/keyword hints from headings (5–8)
10) Positioning statement (fill this: “For [target] who [need], [brand] is a [category] that [unique benefit]. Unlike [alternatives], it [differentiator].”)
11) What they are not saying (notable omissions that could be weak spots)
Return the output as labeled bullets only. Here is the content: [paste homepage hero + pricing + features + about + one case study]

Copy-paste prompt (compare 3–5 competitors)

You are a market analyst. Using the competitor reports above, do three things:
A) Common ground: list the 5–7 claims everyone makes.
B) White space: list 3–5 defendable angles no one (or only one) emphasizes; note buyer value and proof needed.
C) Risk check: where are competitors strongest (proof-rich), and where are they bluffing (claims without evidence)? Keep it tight and actionable.

Copy-paste prompt (draft your positioning)

Based on the white space opportunities identified, write 3 alternative positioning routes. For each route include: 1) Positioning statement, 2) 12-word homepage hero line, 3) Subhead that names the buyer and outcome, 4) 3 proof points I could realistically gather within 60 days, 5) One CTA that reduces risk (trial, audit, template). Keep the language plain and specific.

Worked example (fictitious)

• Competitor A (AcmeCRM)
  • Category: SMB sales CRM with AI forecasting
  • Targets: Sales managers in SaaS, 10–200 seats
  • Pains: Pipeline visibility, rep adoption, forecast accuracy
  • Value + proof: “+22% forecast accuracy”; 3 logo case studies
  • Features: Deal stages, AI scoring, Gmail plugin
  • Pricing: Free, Pro, Enterprise; AI add-on
  • Tone: Confident, numbers-led; CTA: “Start free”
  • Omissions: Weak implementation story
• Competitor B (BrightSales)
  • Category: RevOps platform
  • Targets: RevOps leaders, mid-market
  • Pains: Data silos, reporting
  • Value + proof: “Single source of truth”; vague proof
  • Pricing: Contact sales only
  • Tone: Enterprise, jargon-heavy; CTA: “Book demo”
  • Omissions: No transparent pricing
• Competitor C (CareTrack)
  • Category: Healthcare CRM niche
  • Targets: Clinics; HIPAA first
  • Pains: Compliance, patient follow-up
  • Value + proof: HIPAA badges; 2 healthcare case studies
  • Pricing: Tiered by locations
  • Tone: Trust and safety; CTA: “See compliance checklist”
  • Omissions: Limited AI story

Comparison insight

• Overlap: Everyone claims “visibility” and “centralized data.”
• White space: Fast time-to-value with a 14-day guided setup and guaranteed adoption metric; transparent pricing calculator; compliance + AI story for regulated SMBs.
• Risk: AcmeCRM has evidence on accuracy; BrightSales is light on proof; CareTrack owns compliance.

Common mistakes & quick fixes

• Messy inputs: If AI output feels vague, you likely gave vague inputs. Fix: paste the exact hero, pricing table labels, and one case study quote.
• Over-long reports: Cap each bullet to one line. Ask for a 200–300 word limit.
• Tool blind spots: Some pages block bots. Fix: copy snippets manually or use Reader Mode.
• Shiny object bias: Features ≠ positioning. Always tie features to a buyer outcome and proof.

Action plan (today)

1. List 3 competitors and collect 4–5 URLs each.
2. Run the single-competitor prompt for all three; save results.
3. Run the comparison prompt to spot overlaps and white space.
4. Use the drafting prompt to create 3 positioning routes. Pick one to test.
5. Update your homepage hero and CTA with the chosen route; add or plan proof points.

Expectation setting

• In 30–45 minutes you’ll have standardized snapshots and 2–3 differentiated angles.
• These are hypotheses. Validate fast: a headline A/B test, a pricing page tweak, or a short customer interview.

Closing thought: AI won’t decide your strategy, but it will compress the research time from days to an hour and surface patterns you can act on now.

Good point: Putting the buyer first is exactly the right starting line — realistic, benefit-led descriptions sell more than clever copy aimed at you, the seller.

Why this matters: buyers scan gigs fast. They want to know what they get, how long it takes, and why they should trust you. AI can speed up writing and help you test versions — but you still control the final tone and facts.

What you’ll need

• Clear service name (e.g., “WordPress speed optimization”).
• 3 deliverables (what you’ll deliver).
• Delivery time and revision policy.
• One or two target buyer descriptions (e.g., small business owner, blogger).
• AI tool (ChatGPT or equivalent) and a text editor.

Step-by-step: write a buyer-friendly gig description

1. Gather inputs: title, deliverables, turnaround, price tiers, past results or guarantees.
2. Use a focused AI prompt (copy-paste below) to create 2–3 drafts.
3. Pick the draft that feels most buyer-focused and edit for clarity and truth.
4. Add social proof line (years, clients, outcome) and a simple CTA (e.g., “Message me to get started”).
5. Format for scanning: 1-line benefit, 2 short paragraphs, 3 bullet deliverables, delivery + revisions, CTA.
6. Test: publish, then tweak headline/first line after a week if impressions are low.

Copy-paste AI prompt (use as-is or replace placeholders)

Write a Fiverr gig description (120–160 words) for the service: [SERVICE]. Target buyer: [BUYER PERSONA]. Start with a one-line benefit. Then describe 3 clear deliverables in bullets. State delivery time and revision policy. Include 3 short reasons to hire me and finish with a one-line call-to-action. Tone: friendly, professional, buyer-focused. Include the keywords: [KEYWORDS]. Keep short paragraphs and easy scanning.

Example (before → after)

Before: “I will speed up your WordPress site using caching and image compression.”

After: “Make your website load under 2.5s to keep visitors and boost sales. I’ll deliver: 1) full speed audit with prioritized fixes, 2) caching, image and script optimization, 3) final performance report. Delivery: 3 days • 2 revisions. Why hire me: 5+ years optimizing 200+ sites, measurable speed gains, clear step-by-step report. Ready for faster pages? Message me to get a quick audit.”

Mistakes & fixes

• Too much technical jargon → Fix: translate into buyer outcomes (faster, fewer bounces, more sales).
• Vague deliverables → Fix: list 3 specific items buyers receive.
• No CTA or timeline → Fix: add delivery time and a clear next step.

Action plan (next 48 hours)

1. Write down your 3 deliverables and delivery time.
2. Run the AI prompt once to get 2 drafts.
3. Pick one, tweak for truth, add CTA, and publish.

Small iterative steps win. Use AI to create drafts quickly, but keep the buyer’s perspective front and center — benefits, clarity, and a simple next step will convert better than cleverness alone.

Nice point: Gentle, polite tone is crucial — it keeps relationships intact and improves response rates.

Here’s a practical, do-first guide to using AI to create respectful nudge messages for overdue items. Quick wins you can implement today.

What you’ll need

• Basic info: recipient name, item description, due date, amount (if any).
• An AI tool you can type prompts into (e.g., a chat assistant).
• Templates for different stages: friendly reminder, follow-up, final nudge.
• Channel choice: email, SMS, or app notification.

Step-by-step: create and send your nudges

1. Decide the tone: friendly, polite, short. Keep under 60–90 words for SMS; 80–150 for email preview.
2. Gather variables: name, item, due date, grace period.
3. Ask the AI to draft 3 versions: soft reminder, firm-but-kind follow-up, and final courtesy nudge.
4. Review and edit for accuracy and personalization (one sentence referencing the recipient makes a big difference).
5. Test-send to a colleague or to yourself to check formatting and links.
6. Schedule or send, then track responses and tweak cadence.

Copy-paste AI prompt (use as-is)

“Write three short, polite reminder messages for an overdue library book. 1) Friendly first reminder (gentle tone, under 60 words). 2) Second reminder (firm but courteous, under 80 words). 3) Final courtesy notice (clear next steps, under 100 words). Include placeholders: [Name], [Book Title], [Due Date], [Return Link]. Keep language warm and non-threatening.”

Example outputs (adapt for your use)

• Friendly: “Hi [Name], a quick reminder that [Book Title] was due on [Due Date]. No rush—please return it or let us know if you need more time. Thanks!”
• Second: “Hi [Name], our records show [Book Title] is still checked out since [Due Date]. Please return it or contact us to arrange an extension. We appreciate your help.”
• Final: “Hi [Name], [Book Title] is now overdue. If it’s already returned, thank you—ignore this. If not, please return it or contact us by [date] to avoid penalties.”

Checklist — Do / Don’t

• Do keep messages short and personal.
• Do offer next steps or help options.
• Do test tone with a colleague.
• Don’t use threatening or shaming language.
• Don’t overload with legal or billing detail in the first nudge.

Common mistakes & fixes

• Too formal — Fix: use conversational language and first names.
• Too vague — Fix: include clear next steps and deadlines.
• One-size-fits-all — Fix: tailor templates to customer segment or relationship age.

3-step action plan (today)

1. Use the copy-paste prompt above and generate three messages.
2. Personalize placeholders and test-send one channel.
3. Review replies after 48–72 hours and adjust cadence.

Small, polite nudges work better than loud demands. Start simple, measure, and refine — you’ll get quick wins and keep relationships intact.

Quick win: Paste this two-line prompt into DALL·E and see a usable concept in under 5 minutes — then pick one you like to iterate.

Quick copy-paste prompt (try now):

“Modern minimalist logo, monogram ‘AB’, geometric shapes, flat colors, one accent color #FF6B6B, clean sans-serif feel, centered, transparent background, vector-style, high contrast”

One small correction before we start: DALL·E creates raster images (PNG/JPEG). It can produce clean logo concepts, but not true scalable vector files (SVG) you can edit precisely. Expect to use a vector editor or a trace tool to turn a chosen image into a final, scalable logo.

What you’ll need

• An account with DALL·E or another image-generation tool.
• A short brand brief: one sentence about who you are, one adjective (e.g., modern, warm), and 1–2 color hex codes if you have them.
• Basic image editor or access to a designer to vectorize the chosen concept.

Step-by-step approach

1. Write a short prompt: brand + style + composition + colors + output hints (transparent background, flat colors, centered).
2. Generate 8–12 variations quickly. Look for shapes, balance, and how legible any letters are.
3. Pick 2–3 favorites and refine the prompt: add details like “monoline” or “negative space” or remove clutter.
4. Export the best result. Use a vector trace tool (or designer) to create an SVG and tidy up details like spacing and kerning.

Example prompts — copy, paste, tweak

• Short concept (fast): “Minimalist logo, single icon + wordmark, navy #0A2540 and gold #FFD166, flat, clean lines, transparent background, centered”
• Detailed (more control): “Minimalist logo for a tech consultancy: simple geometric symbol inspired by a compass, monoline, negative space forming a subtle ‘C’, two flat colors (dark teal #0B6E6B and light gray #F2F2F2), no gradients, vector-style, transparent background, balanced spacing”

Common mistakes & fixes

• Asking for precise text or readable small type — Fix: specify placeholder initials and later add final text in a vector editor.
• Too many adjectives = muddled results — Fix: pick 3 max (style, mood, color).
• Expecting vector output — Fix: plan to vectorize the chosen raster image.

Simple action plan (next 30–60 minutes)

1. Create a one-sentence brief and pick 1–2 hex colors.
2. Run the quick-win prompt above and save 5 favorites.
3. Refine top 2 with more detailed prompts and export the best. Then vectorize or hand to a designer.

Closing reminder: Use DALL·E for fast ideas and concepts. Treat the outputs as starting points — then refine, vectorize, and test the logo at different sizes for a professional final result.

Nice focus — you want a mini-course that actually sells on Teachable. That intent is half the battle. AI can speed up content creation, validation and launch — so you get from idea to paying students fast.

What you’ll need

• Clear outcome: one sentence that describes what a student can do after the course.
• Target audience: who they are and their biggest pain.
• Simple tech: Teachable account, Zoom or phone for recordings, a phone or webcam, and basic editing (free tools work).
• AI tool (ChatGPT-like) to draft curriculum, scripts and marketing copy.

Step-by-step plan

1. Validate fast (1–2 days). Post a short poll or ask five people in your network: would they pay $X for outcome Y? Offer a small pre-sell or reservation to test demand.
2. Define the promise + modules (half day). Write a one-line promise and 3–5 module titles with 2–4 micro-lessons each.
3. Use AI to draft the curriculum (1 day). Give AI your audience + promise and ask for learning objectives, lesson titles, activity and a quiz question for each lesson.
4. Turn lessons into short scripts (2–3 days). Use AI to write 5–8 minute lesson scripts, slide bullets and suggested visuals.
5. Create assets (2–4 days). Record videos, export MP4s, create 1-page cheatsheets and upload to Teachable.
6. Pre-launch and price (3–7 days). Offer a launch price to your list or social audience. Use early-bird bonuses to increase urgency.
7. Launch, gather feedback, iterate. Use student feedback to refine content, then raise price gradually.

Copy-paste AI prompt you can use now

“I want to create a 3-module mini-course for [audience: e.g., busy solo entrepreneurs over 40] that promises [outcome: e.g., write high-converting LinkedIn posts in 7 days]. Provide: 1) a course title, 2) 3 module names, 3) 2–3 lesson titles per module with one-line learning objectives, 4) a 1-paragraph course description for sales page, and 5) one short quiz question per lesson.”

Example

• Course: “7-Day LinkedIn Post System for Busy Pros”
• Module 1: Hook & Promise — Lessons: Why posts work, Crafting your hook.
• Module 2: Structure & Value — Lessons: 3-part post structure, Adding credibility.
• Module 3: Publish & Amplify — Lessons: Post schedule, Repurposing content.

Common mistakes & fixes

• Overstuffing content — keep lessons bite-sized (5–10 minutes) and focused.
• Skipping validation — pre-sell or survey before full production.
• Poor pricing — start lower with a clear early-bird limit, then increase.

7-day action plan

1. Day 1: Validate idea with 10 people or a poll.
2. Day 2: Use the AI prompt to generate curriculum.
3. Day 3–5: Create scripts and record lessons.
4. Day 6: Upload to Teachable, build sales page copy using AI.
5. Day 7: Launch pre-sell to your list/social.

Small, consistent steps win. Build, test, improve — and use AI to speed the writing and structure, not replace your voice.

Smart topic. Rapid A/B testing with AI is the fastest way to find winning creatives without burning months or budgets. Let’s turn this into a repeatable, low-stress system you can run every week.

Why this matters

• You don’t need to be a designer or data scientist. AI will generate options, keep tests tidy, and crunch results.
• The goal isn’t perfect creatives; it’s fast learning. Small, disciplined tests compound into big gains.

What you need

• An AI writing assistant (for copy) and an AI image tool (for visuals) — any reputable option works.
• Your ad platform (Meta, Google, LinkedIn) with A/B testing or experiments enabled.
• A simple spreadsheet (Google Sheets or Excel) for naming, test plans, and results.
• Clarity on your single success metric (e.g., cost per lead, purchase ROAS, CTR).

The 7-step quick test system

1. Pick one variable to testChoose only one of: headline, primary text, image/video, or CTA. Keep audience, placement, and budget identical.
2. Map a simple test matrixUse angles, not random ideas. Try the 5P angles: Problem, Proof, Product, Promise, Personality. Plan 2–3 variants per angle (10–15 total).
3. Generate variants with AI (copy + visuals)Feed AI your brand voice and constraints (tone, claims, compliance). See prompt below.
4. Create images fastTurn AI suggestions into images using your preferred tool. Keep formats consistent (e.g., 1080×1080 and 1080×1920). Use simple, high-contrast layouts with one focal point and a readable headline.
5. Name everything clearlyUse a convention like: CAMPAIGN_AUDIENCE_ANGLE_ELEMENT_VERSION. Example: SpringSale_Prospects_Proof_Image_V2. This saves hours later.
6. Set up the platform testUse built-in A/B testing/Experiments. Same audience, same budget, one variable. Run 3–5 days or until each variant hits a minimum sample (see “What to expect”).
7. Export results and analyze with AIPull a CSV with impressions, clicks, spend, conversions. Ask AI to flag winners, estimate uplift, and suggest the next test.

Copy-paste prompt: generate a focused creative test

Use this in your AI writing tool. Replace the [brackets].

• 
  You are a senior performance marketer. Create a rapid A/B test plan for paid [platform] to improve [primary KPI, e.g., cost per lead]. Brand voice: [describe tone, banned words, compliance rules]. Product: [what it does, who it helps, price range]. Audience: [who they are].

  Deliver:
  1) 5 angles (Problem, Proof, Product, Promise, Personality) with 2 headline options and 1 primary text each (under [X] characters where needed).
  2) 3 CTA options.
  3) 3 simple image concepts per angle (describe headline text on image, background, focal object, color, and layout).
  4) A clean naming convention for all variants.
  5) One control ad (best guess) and 10 test variants.
  6) A 5-day test plan with budget split and the single variable to isolate.
  Ensure all claims are supportable and compliant. Keep language clear, everyday, and benefit-first.
  

Copy-paste prompt: quick results analysis

• 
  I’ll paste a table with columns: Variant, Impressions, Clicks, Spend, Conversions. Calculate CTR, CPC, CVR, CPA. Identify the top 2 variants on the primary KPI [e.g., CPA]. Estimate uplift vs control with simple 95% significance guidance (note if sample is insufficient). Recommend what to test next while keeping the winning element constant. Return a 1-paragraph summary and a bullet list of actions.
  

What to expect (practical guardrails)

• Decision sample: Aim for at least 100–300 clicks or 20–40 conversions per variant for early reads. If traffic is low, extend the test.
• Timeframe: 3–7 days per round is common. Avoid judging in the first 24 hours.
• Budget: Keep it even across variants. If unsure, split your normal daily budget evenly for the test window.
• Winner rules: Pre-commit. Example: “Pick any variant with ≥15% improvement on [KPI] and stable trends for 48 hours.”

Example: skincare DTC, testing the image

1. Objective: Lower CPA on Meta.
2. Variable: Image only (keep copy/CTA fixed).
3. Angles: Problem (acne), Proof (before/after), Product (texture/ingredients), Promise (clear skin in 30 days), Personality (founder selfie).
4. AI generates: 10 image concepts with on-image headline text (“Dermatologist-formulated,” etc.).
5. Build: 5 images (square + story size). Name using the convention.
6. Run: 5 variants, even budget, same audience, 5 days.
7. Analyze: AI flags “Proof_BA_V2” with 22% lower CPA. Next round: keep image, test headlines.

Insider tricks

• The ladder method: Round 1: test big concepts (angles). Round 2: test headlines within the winning concept. Round 3: test CTA or first line. This stacks gains without chaos.
• Visual twins: Create a “clean” and a “bold” version of the same layout. One usually wins fast and sets your brand’s tolerance for contrast and text-on-image.
• Evidence beats adjectives: Swap fuzzy claims for numbers or specifics. AI can rewrite: “powerful serum” → “clinically reviewed by 128 customers, average rating 4.7/5.”

Common mistakes and quick fixes

• Too many variables at once → Fix: lock everything but one element.
• Stopping early → Fix: wait for the pre-agreed sample or days-in-market.
• Poor naming → Fix: adopt the convention and use a tracker sheet.
• Changing budgets mid-test → Fix: set and forget until the test ends.
• Ignoring audience overlap → Fix: keep one audience per test or use platform experiments that split traffic cleanly.
• Compliance rejections → Fix: brief AI with banned words and required disclosures upfront.

7-day action sprint

1. Day 1: Pick KPI, choose one variable, draft angles.
2. Day 2: Use the generation prompt to create copy and image concepts.
3. Day 3: Produce visuals and assemble creatives.
4. Day 4: Set up A/B test with clean naming and even budget.
5. Days 5–6: Let it run. No tweaks.
6. Day 7: Export results, run the analysis prompt, lock a winner, and plan the next ladder step.

Final thought

Think “test to learn” not “test to win.” AI makes it easy to ship many small, smart tests. Keep the scope narrow, the names clean, and the cadence weekly. The compounding gains will surprise you.

Quick win: In the next 5 minutes, paste the prompt below into an image generator to create 3 visual concepts. Upload them to a single page in a free prototyping tool and send the link to 5 people for a 2-question reaction — you’ll get usable feedback fast.

Context: AI is excellent for rapid ideation and early user testing of visual concepts. One point to refine: AI doesn’t replace designers or user research — it accelerates the sketching and testing cycle so you can learn faster and iterate smarter.

What you’ll need

• An AI image generator (DALL·E, Midjourney, Stable Diffusion or similar).
• A simple prototyping or presentation tool (Figma, Canva, or PowerPoint).
• A way to collect quick feedback (Google Form, Typeform, or a 1–1 video call).
• 5 people who match your target audience for fast testing.

Step-by-step: how to do it

1. Generate concepts: use the prompt below to create 3 variations (different colors or layouts).
2. Pick the best 3 images and place each on its own frame or slide with a one-line headline and short caption.
3. Turn frames into a shareable link or PDF. Aim for one page per concept.
4. Create a 2-question feedback form: (1) Which concept would you choose and why? (open), (2) Rate confidence 1–5 (quant).
5. Send to 5 target users. Ask them to review in 2–3 minutes and submit responses.
6. Analyze: look for patterns in which concept people prefer and the reasons they give. Iterate the prompt based on insights.

Copy-paste AI prompt (use as-is)

“Create three distinct concept images for a modern portable espresso maker landing page aimed at urban coffee lovers aged 25–45. Concept A: minimalist white and chrome product on a bright kitchen counter, soft natural light, close-up on the product with a cup. Concept B: lifestyle shot of a person pouring espresso in a small apartment, warm tones, cozy mood. Concept C: bold graphic style, high-contrast colors, flat illustration with iconography showing portability and speed. Keep image ratio 16:9, high detail, clean composition, and include space in the top area for a short headline.”

Example

Result: Concept B got the most emotional comments — people mentioned “homey” and “authentic.” That tells you to lean into lifestyle imagery in the next iteration and test price messaging alongside visuals.

Common mistakes & fixes

• Vague prompts → get specific about mood, colors, composition. Fix: specify style, ratio, audience.
• Testing with the wrong people → recruit people matching your audience. Fix: use existing customers or ask contacts who fit the profile.
• Too many elements on one page → keep each concept simple so feedback focuses on the visual idea.

Action plan (do-first timeline)

• 5 minutes: generate 3 concepts with the prompt above.
• 30–60 minutes: assemble into a shareable page and create a 2-question form.
• 1 day: collect feedback from 5 people and decide next iteration.

Reminder: Do the small experiment now. Real learning comes from seeing how real people react, not from waiting for perfect designs. Use AI to create fast drafts, test them, then refine with real-user insight.