The 13-Day Claude Fable Sprint: A Quarter of Ad Creative Strategy Before June 23

TL;DR

The 13-Day Claude Fable Sprint is a step-by-step workflow for Meta advertisers who want to compress a quarter of creative strategy into the window before Claude Fable 5 moves to usage credits on June 23, 2026. Over 13 days, the sprint produces four assets: a full audit of your ad account, a competitor strategy document built from 30 to 40 real ads, an opportunity-scored angle matrix, and 30 production-ready creative briefs. Days 10 through 13 turn those briefs into finished static and video ads. Every prompt and template the sprint requires is written out in full on this page.

Why Does This Window Exist?

Anthropic released Claude Fable 5 on June 9, 2026. It is the company's most capable generally available model and the first public Mythos-class model. Right now it is included in paid Claude plans, meaning Pro, Max, Team, and seat-based Enterprise, at no extra cost through June 22, 2026. From June 23, using it on those plans requires usage credits.

That gives you 13 days of heavy, repeated access to the strongest model Anthropic ships, on a subscription you already pay for. The sprint below is built to spend that access on the work that compounds: strategy, not chat.

One capability matters more than anything else for paid ads. Fable's vision is state of the art, which means it can analyze actual ad creatives from screenshots. You are no longer describing a competitor's ad to a model and hoping your description was fair. You screenshot the ad, send it, and the model reads the hook, the layout, the offer, and the emotional play directly from the pixels. That is what makes the competitor teardown in this sprint possible at this depth.

What You Need Before Day 1

Four things. Nothing else.

• A paid Claude plan: Pro, Max, Team, or seat-based Enterprise. Fable 5 is included on these through June 22, 2026.
• Access to your Meta ad account exports, or at least Ads Manager reporting with ad-level data for the last 90 days.
• Meta Ad Library open in another tab. It is public and requires no login.
• A folder on your desktop for screenshots. You will fill it on Days 3 to 5.

If you have those, you can start Day 1 in the next ten minutes.

Days 1-2: How Do You Audit Your Own Ad Account?

Strategy starts with an honest picture of what you are already running. Most accounts carry fatigued creatives quietly burning budget, and most media buyers know it without being able to point at the exact ads. Day 1 is the export. Day 2 is the diagnosis.

Get the right export

In Ads Manager, set the date range to the last 90 days and the breakdown to ad level, not campaign or ad set level. Then export with these columns: ad name, ad set name, spend, impressions, frequency, CTR (all), CPM, results, and cost per result. If your reporting view supports it, add a weekly breakdown so the model can see decay over time rather than a single averaged number.

Export as CSV and open it. If your account is large, sort by spend and take the top 100 rows. Claude handles pasted tables well, and the top spenders are where the fatigue is hiding anyway.

If you want to skip manual exports entirely, you can wire Claude directly to your ad account through the Meta Ads MCP. We wrote up the full setup in our guide to connecting Claude Code to Meta Ads . For this sprint, the CSV paste works fine.

Run the audit prompt

Paste the prompt below into Claude with Fable 5 selected, fill in the four bracketed blanks, and paste your CSV rows at the end. Copy everything between the markers.

===== COPY FROM HERE: FABLE ACCOUNT AUDIT PROMPT =====

You are a senior creative strategist who has managed Meta ad accounts for ten years. You diagnose creative fatigue from performance data and you are direct about what to kill. You do not soften bad news and you do not pad your analysis with generic advice.

I am pasting an ad-level export from Meta Ads Manager covering the last 90 days. The columns are: ad name, ad set name, spend, impressions, frequency, CTR (all), CPM, results, and cost per result. My product and niche: [YOUR PRODUCT AND NICHE, ONE SENTENCE]. My primary conversion event: [PURCHASE, LEAD, TRIAL, OR INSTALL]. My target cost per result: [YOUR TARGET CPA OR ROAS]. Anything unusual about the flight you should know: [AUDIENCE CHANGES, PROMOS, OR SEASONALITY DURING THE PERIOD, OR WRITE NONE].

Use these fatigue criteria, exactly. Flag a creative as fatigued if at least two of the following are true: one, frequency is above 3.0 while its CTR sits below the account median. Two, CTR has declined more than 30 percent compared with its first two weeks of delivery, if weekly data is present. Three, CPM has inflated more than 20 percent over the flight while the targeting stayed the same. Separately, flag a creative as a confirmed loser if it has spent more than three times my target cost per result without producing a conversion.

Return four things in this order. First, a table of fatigued creatives with columns: ad name, spend, the specific criteria it triggered, and a one-sentence diagnosis. Second, a ranked kill list, worst offender first, with the monthly spend I would recover by pausing each one. Third, the patterns across my winners: what the top 20 percent of ads by cost per result share in format, hook, and angle, written as testable hypotheses rather than vague observations. Fourth, three questions about gaps in the data that would change your analysis if answered.

Here is the export: [PASTE YOUR CSV ROWS HERE]

===== COPY TO HERE =====

What good output looks like

A strong audit names specific ads with specific numbers: this ad triggered criteria one and three, frequency 4.2 with CTR 22 percent under median, CPM up 31 percent since week two, pause it and recover roughly $1,800 a month. If the output reads like a blog post about creative fatigue in general, the data did not come through cleanly. Repaste it as plain rows and ask again.

One tip for the second pass: do not accept the first answer. Reply with a single line, such as: which item on your kill list are you least confident about, and what evidence would change your mind? Fable is at its best when you make it argue against itself. The kill list that survives that exchange is the one you act on.

Days 3-5: How Do You Tear Down Your Competitors' Ads?

This is the step Fable's vision was made for, and the part of the sprint with the highest payoff per hour. You are going to capture 30 to 40 live competitor ads and have the model read the strategy behind every one of them.

Collect the ads

1. Open Meta Ad Library, set your country, and search each competitor by brand name. Filter to active ads.
2. Pick three to five competitors. Direct competitors first, then one or two brands a tier above you whose creative you admire.
3. Capture 30 to 40 ads total. Enough for real patterns, small enough to finish in a day.
4. Screenshot each ad cleanly: the full ad card including primary text, headline, and creative, with the browser chrome cropped out. Capture the start date too, because longevity usually means an ad is working.
5. Name files by competitor and number, like brandA-01.png, so you can reference them later.

Batch them into Claude

Send the screenshots in groups of 8 to 10 per message rather than all at once. One message can only carry so many images, and attention per ad drops when a single batch is overloaded. Paste the prompt below with your first batch, then send the remaining batches with one line each: same framework, ads 11 to 20 from [COMPETITOR]. After the last batch, send the single word SYNTHESIZE.

===== COPY FROM HERE: FABLE COMPETITOR TEARDOWN PROMPT =====

You are a competitive creative strategist for paid social. You have analyzed thousands of Meta ads and you can read an ad's strategy from the creative alone: the hook, the awareness level it targets, the offer mechanics, and the emotion doing the selling.

I am sending you [TOTAL NUMBER] screenshots of active Meta ads from my competitors in [YOUR NICHE]. They will arrive in batches of 8 to 10. Analyze every ad individually before any synthesis. Number each ad in the order received and keep the numbering consistent across batches.

For each ad, give me one row in a table with these columns. Ad number and competitor. Hook type: curiosity, problem call-out, social proof, statistic, before-and-after, question, or pattern interrupt. Angle: the core persuasive argument in five words or fewer. Awareness level: unaware, problem aware, solution aware, product aware, or most aware. Offer structure: discount, free trial, bundle, guarantee, free shipping, or none visible. Format: static, UGC video, studio video, carousel, or motion graphic. Emotional driver: fear, status, belonging, relief, aspiration, anger, or curiosity.

Do not synthesize until I send the word SYNTHESIZE. When I do, produce a strategy document with this exact structure. Section 1, their dominant angles: the three angles this market spends the most creative volume on, with the competitors running each and what that concentration tells us. Section 2, their gaps: the angles and awareness levels nobody is covering, with your hypothesis for why each gap exists, whether it is an oversight or a tested dead end. Section 3, the white space: three angles nobody in this niche is running that fit my product, ranked by how defensible each would be if competitors copied it. Section 4, format and offer patterns: which formats dominate, which offers repeat, and where the market has converged so hard that simply being different is cheap attention.

My product, for the white space section: [ONE SENTENCE ON YOUR PRODUCT AND WHO IT IS FOR].

===== COPY TO HERE =====

The destination format

Here is the skeleton of the strategy document you should end Day 5 holding. If the synthesis is missing a piece, ask for it by name. Save the finished version as a text file, because it is an input for Day 6.

===== COPY FROM HERE: COMPETITOR STRATEGY DOC TEMPLATE =====

COMPETITOR CREATIVE STRATEGY: [NICHE], [DATE]

ADS ANALYZED: [NUMBER] active ads across [COMPETITOR NAMES]

1. DOMINANT ANGLES. Angle A: [argument], run by [competitors], [share of ads analyzed]. Angle B: [argument], run by [competitors]. Angle C: [argument], run by [competitors]. What the concentration means: [two sentences].

2. GAPS. Uncovered angle or awareness level 1: [description], likely reason: [hypothesis]. Uncovered angle or awareness level 2: [description], likely reason: [hypothesis]. Uncovered angle or awareness level 3: [description], likely reason: [hypothesis].

3. WHITE SPACE. Opportunity 1: [angle], defensibility: [high, medium, or low, with reason]. Opportunity 2: [angle], defensibility: [rating and reason]. Opportunity 3: [angle], defensibility: [rating and reason].

4. FORMAT AND OFFER PATTERNS. Dominant formats: [breakdown]. Repeating offers: [list]. Convergence note: [where everyone looks identical and what a cheap differentiation would be].

===== COPY TO HERE =====

Days 6-9: How Do You Turn Research Into 30 Briefs?

You now hold two documents: your account audit and the competitor strategy doc. Days 6 and 7 fuse them into a scored testing plan. Days 8 and 9 expand that plan into 30 briefs a production team can execute without asking you a single question.

First prompt: the angle matrix

===== COPY FROM HERE: FABLE ANGLE MATRIX PROMPT =====

You are a creative strategist planning a quarter of Meta ad testing. You think in portfolios, not single ads: you spread bets across angles, personas, and awareness levels, weighted by the evidence behind each one.

I am pasting two documents. Document one is an audit of my own ad account: my winners, my losers, and the patterns across them. Document two is a competitor strategy document: the market's dominant angles, its gaps, and the white space.

Build me an angle matrix. The rows are every angle worth testing, drawn from three sources: my proven winners that deserve fresh executions, the competitor gaps, and the white space opportunities. The columns are persona and awareness level. My personas: [LIST YOUR TWO TO FOUR PERSONAS IN ONE LINE EACH, OR WRITE: PROPOSE THEM FROM THE DOCUMENTS].

For each angle-persona-awareness combination that makes strategic sense, score the opportunity from 1 to 10 using three factors: evidence it can work, from my data or from sustained competitor spend; competition density, where gaps and white space score higher; and fit with my product's actual strengths. Show the score and a one-line rationale. Skip combinations that make no sense rather than forcing a score.

Then give me the top 10 combinations ranked by score, and flag any cell where you are guessing rather than reasoning from the two documents. I want to know which scores to trust.

Document one, my account audit: [PASTE YOUR AUDIT OUTPUT]. Document two, the competitor strategy doc: [PASTE YOUR STRATEGY DOC].

===== COPY TO HERE =====

Sanity-check the top 10 before moving on. If a combination scored 9 but your gut disagrees, say so and make the model defend the score. Day 7 exists for that argument. The matrix you carry into Day 8 should be one you actually believe.

Second prompt: 30 production-ready briefs

===== COPY FROM HERE: FABLE BRIEF GENERATION PROMPT =====

You are a senior creative strategist writing production briefs for a designer and editor who have never spoken to me and cannot ask follow-up questions. Every brief must be executable cold.

I am pasting my scored angle matrix. Generate 30 briefs drawn from the highest-scoring combinations. Spread the formats roughly as: [12] static image, [12] UGC-style video, and [6] motion graphic. Adjust the split when an angle clearly demands a specific format, and tell me each time you do.

Use the exact template I provide below for every brief. Every field filled in, no field skipped, no field answered with a vague phrase. Hooks must be written out word for word, not described. Visual direction must be specific enough that two different designers would produce recognizably similar ads.

Number the briefs S01 through S30. After the final brief, list the five you would launch first if I could only produce five this week, with one line of reasoning each.

The template: [PASTE THE BRIEF TEMPLATE FROM BELOW]. The matrix: [PASTE YOUR ANGLE MATRIX OUTPUT].

===== COPY TO HERE =====

The brief template

This is the template the prompt above references. Paste it in where the placeholder indicates. The fields map one-to-one onto production inputs, which matters on Day 10.

===== COPY FROM HERE: CREATIVE BRIEF TEMPLATE =====

BRIEF ID: [S01 through S30]

ANGLE: [the persuasive argument this ad makes, one sentence]

PERSONA: [who this ad speaks to, in their own words]

AWARENESS LEVEL: [unaware, problem aware, solution aware, product aware, or most aware]

FORMAT: [static, UGC video, or motion graphic]

HOOK OPTION 1: [exact first line or opening three seconds, written out]

HOOK OPTION 2: [exact alternative hook, written out]

HOOK OPTION 3: [exact alternative hook, written out]

VISUAL DIRECTION: [two to four sentences a designer can execute: scene, subject, composition, where text overlays sit]

COPY DIRECTION: [the argument the primary text makes, the proof it leans on, and the tone]

CTA: [button text plus the closing line of copy]

REFERENCE AD: [ad number from the competitor teardown, or NONE]

SUCCESS METRIC: [the single number that decides if this brief wins, such as hook rate above 30 percent or CPA below target]

===== COPY TO HERE =====

End of Day 9, read five briefs at random. If you can picture the finished ad from the brief alone, the document is production grade. If you cannot, send the weak ones back with one line on what is missing.

Days 10-13: How Do 30 Briefs Become Finished Ads?

Here is the honest part. Thirty briefs is where most teams stall. A brief without production behind it is a document, and documents do not get impressions. From Day 10 the bottleneck is volume, and you solve volume by batching production by format instead of going brief by brief.

This is where we use HeyOz, which hosts image and video models like GPT Image 2 and Seedance 2.0 in one place. It is model-agnostic production infrastructure, so the briefs Fable wrote drop straight in. Fable does the thinking, HeyOz does the making.

Day 10: all twelve statics

Work through every static brief in one sitting. Take each brief's VISUAL DIRECTION field and HOOK OPTION 1 into HeyOz and generate with GPT Image 2: the visual direction becomes the image prompt, the hook becomes the text overlay. Two or three executions per brief, pick the strongest, move on. Twelve briefs at this pace is an afternoon.

Days 11-12: video variants

Batch the twelve UGC video briefs next, generating with Seedance 2.0. The PERSONA field tells you who appears on screen, the three HOOK OPTIONS give you three opening variants per brief, and the COPY DIRECTION drives the script. If you want a frame-locked process that keeps characters consistent across shots, we wrote the full method in the storyboard method for AI video ads , and a volume-focused companion in our guide to creating AI UGC ads at scale . The six motion graphic briefs close out Day 12 with the same inputs.

Day 13: QA and handoff

Check every finished ad against its brief's SUCCESS METRIC field and name every file by its BRIEF ID, so any winner traces back to the angle, persona, and awareness level that produced it. That naming discipline turns next month's data into next quarter's strategy. Upload, schedule your launches, and the sprint is done.

What Happens After June 23?

The workflow does not expire. Every prompt on this page runs on Claude Opus or Claude Sonnet after the window closes, and the 13-day structure holds regardless of which model executes it. If you are reading this in July or later, run the sprint as written with the model your plan includes.

What changes is the economics of the heaviest step. Fable's vision and long-session reasoning made it sharpest for the competitor teardown, where it holds 40 ads in one analytical thread without losing the numbering. After June 22 it requires usage credits, so the question becomes whether the teardown is worth the credits to you. For an account spending real money on Meta, it usually is. For everything else in the sprint, the included models do the job well.

Frequently Asked Questions

Can Claude analyze Meta ads?

Yes. Claude analyzes Meta ads in two ways: pasted performance data from Ads Manager exports, and screenshots of ad creatives. Claude Fable 5's vision capabilities are state of the art, so it can read hooks, offers, formats, and emotional drivers directly from screenshots of real ads.

Is Claude Fable 5 free?

No. Claude Fable 5 is included in paid Claude plans, meaning Pro, Max, Team, and seat-based Enterprise, at no extra cost through June 22, 2026. From June 23, 2026, it requires usage credits on those plans. It was released on June 9, 2026.

What is Claude Fable 5?

Claude Fable 5 is Anthropic's most capable generally available model and the first public Mythos-class model, released on June 9, 2026. For advertisers, its most relevant strength is vision: it can analyze actual ad creatives from screenshots rather than working from text descriptions.

How do I get my ad data into Claude?

Export an ad-level report from Meta Ads Manager covering the last 90 days, with spend, impressions, frequency, CTR, CPM, results, and cost per result as columns. Save it as CSV and paste the rows directly into the chat. For large accounts, paste the top 100 ads by spend.

How many ad creatives should I test per month?

Most performance teams testing seriously on Meta refresh 20 to 40 creatives per month, scaled to spend. The practical constraint is rarely ideas, it is production volume, which is why this sprint pairs 30 briefs with a batched production process rather than stopping at strategy.

Does this workflow work for Google or TikTok ads?

The structure transfers: audit your account, tear down competitors, build an angle matrix, write briefs, produce in batches. The specifics on this page are Meta-native, since the export columns and the Ad Library steps assume Meta's tooling. Swap in your platform's reporting and creative library and the prompts need only light edits.

Can I run the sprint in fewer than 13 days?

Yes. The 13 days assume a normal workload alongside the sprint. With a clear calendar, the strategy phase compresses to four or five days. The sequence matters more than the pace: audit before teardown, teardown before matrix, matrix before briefs.

Start Day 1

The sprint begins with a 90-day export, and the window ends June 22. Run Days 1 and 2 today and you will know more about your account by tonight than most audits deliver in a week.

And when your 30 briefs are ready, the production half runs in one place. HeyOz hosts GPT Image 2 and Seedance 2.0 natively, so the briefs become finished statics and videos without carrying files between tools. Start your 3-day free trial and have Day 10 covered before you get there.