Fix Weak Creative Insight with ChatGPT-Driven Ad Analysis
Marketing teams sit on mountains of ad variants and performance data but still can’t say which words, images or formats truly move ROAS. This article shows how to use ChatGPT to dissect creatives at scale, uncover patterns behind clicks and conversions, and turn insights into a faster, smarter testing engine for your ads.
Inhalt
The Challenge: Weak Creative Performance Insight
Modern marketing teams run hundreds of ad variants across Meta, Google, TikTok, LinkedIn and display networks. Yet when performance drops or CPAs creep up, it is surprisingly hard to answer a simple question: which specific creative elements are actually driving results? Headlines, hooks, CTAs, colors, layouts, emotional angles and value propositions are all mixed together, making it nearly impossible to isolate what works and what doesn’t.
Traditional approaches rely on manual reporting in spreadsheets, occasional deep dives from an analyst, or relying on the limited breakdowns inside each ad platform. These methods break down once you have dozens of campaigns and hundreds of assets. Marketers spend hours tagging screenshots, exporting CSVs, and trying to eyeball patterns across channels. By the time a conclusion is reached, the auction dynamics and audience behavior have already shifted. Manual analysis is too slow and too shallow for today’s creative testing velocity.
The impact on the business is real. Without clear creative insight, budgets continue to flow into underperforming formats, while high-potential angles are underfunded or even turned off. CPAs rise, ROAS erodes, and teams fall back on generic creative that feels safe but doesn’t differentiate. Internally, discussions between brand, performance and leadership become opinion-driven instead of data-driven, slowing decisions and making it harder to justify spend or push bold creative bets.
The good news: this problem is highly solvable. Advances in generative AI and language models such as ChatGPT make it possible to systematically analyze copy, visuals and performance data in one place—and at a speed no human team can match. At Reruption, we’ve seen how an AI-first lens on marketing workflows can turn messy creative data into a clear testing strategy in a matter of days. In the rest of this guide, we’ll walk through practical ways to use ChatGPT to decode your creatives and build a repeatable, insight-driven optimization loop.
Build an AI system with us now!
We build a proof of concept for your problem for 5,000–8,000€. You get a tangible demo instead of slides with promises.
Innovators at these companies trust us:
Our Assessment
A strategic assessment of the challenge and high-level tips how to tackle it.
From Reruption’s work building AI-first capabilities inside marketing and commercial teams, we’ve seen a consistent pattern: the biggest gains do not come from yet another reporting dashboard, but from changing how teams make decisions. Used correctly, ChatGPT becomes an analyst and creative strategist in one—able to read your copy, interpret visual descriptions or image uploads, and connect them to performance data to surface patterns your team can actually act on.
Treat ChatGPT as a Creative Intelligence Layer, Not Just a Copy Tool
Most teams meet ChatGPT first as a text generator: write some headlines, draft ad copy, translate assets. That’s useful, but it misses the real leverage. The strategic shift is to position ChatGPT as a creative intelligence layer across your ad stack. It should help you understand why an ad works, not just produce more variants.
This means feeding ChatGPT both qualitative and quantitative inputs: performance metrics, targeting context, brand positioning, and raw creatives. With this view, it can cluster patterns (“short benefit-led headlines with explicit pricing outperform emotional storytelling for remarketing audiences”) instead of making generic suggestions. Strategically, you move from “AI for more content” to AI for better decisions about content.
Start With a Narrow, High-Impact Use Case
To avoid overwhelm and stakeholder skepticism, resist the urge to “AI-ify” the entire funnel at once. Instead, identify one narrow but impactful scenario where weak creative performance insight is clearly hurting ROAS: for example, Meta prospecting campaigns in one region, or Google Performance Max creatives for a key product line.
Concentrating on a single slice allows you to define clean inputs (a specific export from your ad platform), clear metrics (e.g. CTR, CVR, ROAS) and fast feedback. You can then demonstrate within weeks how ChatGPT-powered analysis leads to better creative decisions and improved performance, before scaling the approach to other channels and markets.
Prepare Your Data and Taxonomy Before You Scale
ChatGPT can work with messy inputs, but you will get far more strategic value if you establish a basic creative taxonomy and data structure. For example, tag assets with dimensions such as offer type, angle (social proof, urgency, savings), format (UGC, product-only, lifestyle), and main visual element. Even simple, consistent tags dramatically improve the patterns ChatGPT can surface.
At an organizational level, this means aligning brand, performance marketing and analytics on a shared language for creative elements. With Reruption’s AI engineering experience, we often help teams automate part of this tagging using computer vision or rule-based scripts, then feed the structured data into ChatGPT. Strategically, this preparation step unlocks scalable cross-channel creative insight instead of one-off analyses.
Design a Human-in-the-Loop Review Process
Even the best AI-driven creative analysis should not run unchecked. You need a clear process where performance marketers and brand owners review ChatGPT’s hypotheses, validate them against their own understanding, and decide which ideas enter the test roadmap. This protects brand integrity and avoids overreacting to short-term data noise.
Strategically, position ChatGPT as augmenting your team’s judgment, not replacing it. Make it explicit who is responsible for accepting or rejecting AI-suggested test ideas, how often insights are reviewed (e.g. weekly creative review), and how learnings are documented. This human-in-the-loop setup builds trust and ensures that AI output translates into consistent creative improvements.
Manage Risk Through Guardrails and Governance
Introducing AI into your marketing decision-making also introduces new risks: overfitting to short time periods, misinterpreting causality, or accidentally drifting from brand and compliance guidelines. You need governance around what data is shared with ChatGPT, which use cases are allowed, and how outputs are checked.
From a strategic perspective, define clear guardrails: no use of first-party PII in prompts, no automatic activation of campaigns based solely on AI analysis, and explicit brand tone and compliance rules embedded in every workflow. Reruption’s work across AI Strategy, Security & Compliance shows that getting these basics right early allows teams to scale safe, compliant AI usage in marketing without constant firefighting later.
Using ChatGPT for creative performance insight is ultimately about turning scattered ad data into a systematic learning engine. When you combine structured inputs, a focused scope, and human-in-the-loop governance, ChatGPT can quickly reveal which hooks, visuals and formats truly move ROAS—and help your team test smarter, not just faster. If you want to validate what this could look like in your own marketing setup, Reruption can help you design and implement a tailored AI-powered analysis flow, from a focused PoC to an embedded capability inside your team.
Build an AI system with us now!
We build a proof of concept for your problem for 5,000–8,000€. You get a tangible demo instead of slides with promises.
Real-World Case Studies
From Pharmaceuticals to Payments: Learn how companies successfully use ChatGPT.
Best Practices
Successful implementations follow proven patterns. Have a look at our tactical advice to get started.
Build a Standard Creative Analysis Prompt Template
Consistent prompts lead to consistent insight. Create a reusable template that your team uses every time they ask ChatGPT to analyze ad creatives. Include: channel context, audience, campaign objective, performance metrics, and a structured list of ads (headline, body, visual description or image, CTA, and results).
Here is a practical starting point you can adapt to your stack:
Act as a senior performance marketing analyst.
Goal: Analyze which creative elements drive performance in my ads and propose specific hypotheses and test ideas.
Context:
- Channel: Meta Ads (Facebook + Instagram)
- Objective: Purchases
- Target audience: [brief audience description]
- Primary KPI: ROAS; Secondary: CTR, CPC
Data format for each ad:
- Ad ID
- Headline
- Primary text
- Visual description (or image reference)
- CTA button
- Impressions
- Clicks
- Spend
- Conversions
- Revenue
Tasks:
1) Group ads into clusters based on creative similarities.
2) For each cluster, identify patterns in copy and visuals tied to strong or weak performance.
3) Formulate 5-10 clear hypotheses (IF > THEN) about what drives performance.
4) Suggest 5 new ad concepts to test based on top-performing patterns.
5) Flag 5 underperforming patterns we should avoid or rework.Expected outcome: a structured summary of winning and losing creative elements that can directly feed into your next creative briefing or sprint.
Use Image Input to Decode Visual Patterns
Text alone often misses what truly differentiates your creatives. With image input, ChatGPT can analyze your ad visuals directly. Upload a selection of winning and underperforming image or video thumbnails, plus their performance metrics, and ask ChatGPT to describe and compare them.
Example workflow:
You are a creative analyst for paid social.
I will upload several ad images. For each, I will also provide key performance metrics.
Tasks:
1) Describe each image precisely (composition, colors, product visibility, people, text in image, style: UGC vs studio, etc.).
2) Compare high-performing vs low-performing images and identify recurring visual patterns.
3) Suggest 5 concrete visual guidelines we should follow in the next shoot or design round.
4) Propose 5 new visual concepts based on the highest-performing patterns.This turns vague feedback like “lifestyle images seem to work” into concrete rules such as “close-up shots of the product in use with a single human subject and high color contrast consistently outperform abstract product-only shots.”
Automate Creative Tagging and Clustering With ChatGPT
Before you can get deep insight, you need structured data. Use ChatGPT to auto-tag creatives with themes, angles and formats. Export your ads (or a subset), paste them into ChatGPT, and have it assign standardized tags that your team agrees on.
Example prompt:
Act as a marketing data analyst.
I will provide a list of ads with the following fields:
Ad ID | Headline | Primary text | Visual description | CTA
For each ad, output a table with:
- Ad ID
- Angle (choose one: discount/savings, social proof, urgency, problem/solution, product benefits, brand story)
- Emotional tone (choose up to 2: rational, aspirational, fear of loss, excitement, trust)
- Format (UGC-style, studio/product-only, lifestyle, graphic/illustration, testimonial)
- Key promise (short phrase summarizing the main promise)
Then, summarize how many ads fall into each category and which categories seem under-tested.Once you have this tagged data, you can run further analysis in ChatGPT or your BI tools to understand which angles or formats drive performance—and where you have blind spots.
Translate Insights into a Structured Testing Roadmap
Analysis without execution does not move ROAS. Use ChatGPT to convert insights into a prioritized test roadmap that slots neatly into your existing sprint or campaign planning. Feed it your constraints (design bandwidth, budget, number of variants you can test per week) and have it build a realistic plan.
Example prompt:
Based on the following creative insights and hypotheses [paste summarized insights],
create a 4-week testing roadmap for our Meta and Google Ads.
Constraints:
- We can produce 6 new creatives per week.
- We can run 4 parallel A/B tests at any time.
- Focus on maximizing ROAS while maintaining brand guidelines [summarize brand constraints].
Output:
- Week-by-week table of tests (channel, audience, hypothesis, creative concept, KPIs to track).
- Clear success criteria for each test.
- A brief summary of how we will turn results into updated creative guidelines.This ensures your team doesn’t just “learn interesting things” from ChatGPT, but systematically turns them into new ads and persistent playbooks.
Create Channel-Specific Insight Summaries for Stakeholders
Senior stakeholders don’t have time to read raw analysis exports. Use ChatGPT to generate concise, channel-specific creative reports that explain what’s working and what will change next. This improves alignment between performance, brand and leadership.
Example workflow:
Act as a marketing insights lead.
I will provide you with:
1) A summary of creative performance insights from our analysis.
2) The list of tests we ran in the last 4 weeks and the results.
Audience: CMO and Head of Brand.
Tasks:
- Create a 1-page executive summary for Meta Ads and a 1-page summary for Google Ads.
- For each, explain in simple language:
- Top 3 learnings about creative performance.
- What we will do differently in the next 4 weeks.
- Any brand-relevant implications.
- Keep it concise and avoid technical jargon.This not only saves time, it also reinforces a culture of evidence-based creative decisions instead of opinion battles.
Integrate AI Analysis Into a Regular Cadence
The real value emerges when ChatGPT-based creative analysis becomes a recurring ritual. Define a regular cadence—weekly for high-spend accounts, bi-weekly for smaller budgets—where you refresh the data, run your standard analysis prompts, and update your testing roadmap.
Operationally, create a simple checklist: export latest performance data, update creative list with new assets, run the tagging prompt, run the analysis prompt, then review insights in a short team session. Over time, this rhythm will dramatically increase the number of validated creative learnings your team accumulates.
Expected outcomes: within 4–8 weeks, most teams can expect clearer creative guidelines, a measurable reduction in wasted spend on low-performing concepts, and incremental ROAS improvements in the 10–25% range on the campaigns where AI-driven insights are consistently applied. Exact numbers depend on baseline performance and execution discipline, but the pattern—fewer guesses, more validated creative decisions—is consistent.
Build an AI system with us now!
We build a proof of concept for your problem for 5,000–8,000€. You get a tangible demo instead of slides with promises.
Frequently Asked Questions
ChatGPT can act as an analytical layer across your creatives and performance data. By feeding it structured data (headlines, body copy, visual descriptions or images, CTAs, targeting context, and metrics like CTR, CVR, ROAS), it can cluster similar ads, compare winners vs. losers, and articulate clear patterns such as “short, benefit-led headlines with explicit pricing outperform emotional stories for cold audiences.”
Instead of scanning endless spreadsheets, your team gets synthesized insights: which angles, formats, and visual styles tend to win, which ones consistently underperform, and which hypotheses to test next. This turns weak, anecdotal creative understanding into a repeatable, data-backed learning process.
You don’t need a large data science team to start. The core requirements are:
- A performance marketer or analyst who can export campaign data from your ad platforms.
- Basic spreadsheet skills to clean and structure the data (e.g. mapping metrics to creatives).
- Someone who understands your brand and target audience to validate ChatGPT’s conclusions.
From a skills perspective, prompt design and basic data structuring are more important than advanced AI knowledge. Over time, you can involve engineering to automate exports and tagging. Reruption often helps teams build this bridge: starting with manual workflows and evolving towards integrated, semi-automated analysis pipelines.
Most teams can see actionable insights within the first 1–2 weeks if they start with a focused campaign set (e.g. one major channel and product line). In the first sessions, ChatGPT will already highlight obvious winning patterns and under-tested angles.
Measurable performance lift typically comes after a full test cycle. For many advertisers, this means 4–8 weeks to design new creatives based on AI insights, run controlled tests, and roll out winners. The key is consistency: integrating AI analysis into your regular optimization cadence rather than treating it as a one-off project.
The direct cost of using ChatGPT for creative analysis is relatively low compared to media spend. The main investment is team time to prepare data, run analyses, and implement findings. Even a few hours per week can be enough to start.
On the ROI side, the upside usually comes from three areas:
- Reduced wasted spend on consistently underperforming creative patterns.
- Higher ROAS on campaigns where winning angles are identified and scaled faster.
- Lower creative production waste because briefs are guided by evidence instead of guesswork.
For many advertisers, a small percentage improvement in ROAS on a key channel already covers the effort many times over. We typically encourage teams to track baseline KPIs and then compare performance for campaigns using AI-driven insights vs. control campaigns.
Reruption combines AI Strategy, AI Engineering, and hands-on marketing workflows to move beyond theory. We can help you in three concrete ways:
- AI PoC for creative insight (9.900€): In a focused Proof of Concept, we define a specific use case (e.g. Meta prospecting campaigns), build a working prototype of a ChatGPT-based analysis flow, and evaluate performance (speed, quality of insights, impact on test outcomes). You get a live demo, metrics, and a production roadmap.
- Embedded implementation support: With our Co-Preneur approach, we work inside your team’s P&L, not just in slide decks. We design prompts, data flows, and processes together with your marketers, and stay until a real solution is shipping and used.
- Enablement and governance: We help you set up guardrails, templates, and a repeatable cadence so that your team can run AI-driven creative analysis safely and independently over time.
If you want to test whether ChatGPT can materially improve your creative performance insight, starting with a scoped PoC is often the fastest and lowest-risk way to get from idea to working solution.
Contact Us!
Contact Directly
Philipp M. W. Hoffmann
Founder & Partner
Address
Reruption GmbH
Falkertstraße 2
70176 Stuttgart
Contact
Phone