Inside the Prompt

Welcome Back to XcessAI

A few days ago, OpenAI released something very interesting: a detailed breakdown of how people use ChatGPT across more than a million conversations.

This kind of transparency is very useful, as it shows investors, enterprises, and the public where generative AI has real traction, and where adoption is thinner than the hype suggests.

These patterns don’t just reflect curiosity. They reveal where users are finding real value, where trust is being built, and where many business opportunities are likely to emerge.

Quick Read

Bottom-line: people lean on AI most for structured guidance and writing tasks, less for complex analytics or emotional support.

• Practical guidance (28%) — tutoring, teaching, and “how-to” advice dominate.

• Writing (28%) — editing, critiquing, summarizing, personal communication are massive use cases.

• Technical help (7.5%) — programming, data analysis remain steady but not dominant.

• Self-expression (4%) — roleplay, reflection, chitchat are small but persistent.

• Data-analysis (0.4%) — surprisingly low, far below what many people might expect.

The Full Breakdown of ChatGPT Usage

OpenAI’s sample of 1.1 million conversations (May 2024–June 2025) shows how people actually use ChatGPT across categories:

• Practical Guidance (28.3%)

  • Tutoring or Teaching — 10.2%

  • How-to Advice — 8.5%

  • Health, Fitness, Beauty & Self-Care — 5.7%

  • Creative Ideation — 3.9%

• Writing (28.1%)

  • Edit or Critique Provided Text — 10.6%

  • Personal Writing / Communication — 8.0%

  • Translation — 4.5%

  • Argument or Summary Generation — 3.6%

  • Write Fiction — 1.4%

• Seeking Information (21.3%)

  • Specific Info — 18.3%

  • Purchasable Products — 2.1%

  • Cooking & Recipes — 0.9%

• Technical Help (7.5%)

  • Computer Programming — 4.2%

  • Mathematical Calculation — 3.0%

  • Data Analysis — 0.4%

• Multimedia (6.0%)

  • Create an Image — 4.2%

  • Generate or Retrieve Other Media — 1.1%

  • Analyse an Image — 0.6%

• Self-Expression (4.3%)

  • Relationships & Personal Reflection — 1.9%

  • Greetings & Chitchat — 2.0%

  • Games & Role Play — 0.4%

• Other / Unknown (4.6%)

Why OpenAI Shared This

We believe that releasing this data serves several purposes:

• Transparency: signaling to regulators and the public that ChatGPT is being used in mainstream, constructive ways.

• Market shaping: highlighting the stickiest categories (education, writing, workflows) to encourage developers and partners to build around them.

• Expectation setting: countering the hype that AI is everywhere, and showing instead that adoption clusters in specific, high-value domains.

• Demand signaling: reinforcing that generative AI is not a fad — millions are already relying on it daily.

• Ecosystem guidance: pointing entrepreneurs and corporates toward the verticals where real user pull already exists (tutoring, health, productivity).

The Usage Patterns That Stand Out

1. Education & Guidance (28.3%)
Tutoring (10.2%) and “how-to” advice (8.5%) together form the single largest slice. People clearly want AI to explain, simplify, and guide them through tasks. This isn’t just students — it’s employees asking for process help, consumers troubleshooting, and professionals brushing up on skills.

2. Writing (28.1%)
Editing and critique (10.6%), personal communication (8%), summaries (3.6%) — writing dominates. This validates the “AI as copilot” vision: embedding LLMs inside email, contracts, PR, HR, and legal workflows.

3. Health & Self-Care (5.7%)
A smaller but telling slice. People already trust AI with personal advice on fitness, nutrition, or well-being. Adoption is early, but intent is strong.

4. Technical Help (7.5%)
Programming (4.2%), math (3%), data analysis (0.4%). The real surprise here is how tiny data analysis is — just 0.4% of usage. For a category executives often assume AI is transforming, the reality is that adoption barely registers. This suggests either current tools aren’t delivering, or users still prefer specialized platforms for analytics.

5. Self-Expression (4.3%)
Roleplay, reflection, greetings. This isn’t a trivial use case — it signals AI’s stickiness in companionship and creativity, even if small in percentage.

What’s Missing

Some gaps are as important as the highlights:

• Data analysis (0.4%) — very low share compared to its importance in business.

• Purchasable products (2.1%) — small slice, but direct monetization potential is huge.

• Cooking & recipes (0.9%) — surprisingly niche despite broad consumer interest.

These may be early indicators of areas where AI hasn’t yet found the right product fit.

Implications for Business

• Productivity is clustering. Employees lean on AI for writing, guidance, and lightweight technical tasks — less for deep analytics.

• Education pull is strong. Whether in corporate training, compliance, or consumer self-learning, AI tutoring has organic traction.

• Trust will matter. Categories like health, shopping, and translation already exist — but winning them requires credibility, context, and localization.

• Blind spots remain. Analytics, domain-specific tasks, and decision-making are still underserved — opportunity space for targeted solutions.

The Opportunity Wedges

Every bar in this chart corresponds to behaviour people already pay for elsewhere:

• Tutoring & teaching → corporate training, compliance, personalized learning.

• How-to advice → vertical AI specialists (legal paperwork, job-hunting, e-commerce setup).

• Writing copilots → embedding into enterprise workflows (legal, HR, PR).

• Health & self-care → hybrid AI + human coaching.

• AI shopping layer → the rebirth of affiliate marketing, but AI-first.

• Translation & localization → tools that translate intent, not just words.

• Vertical code copilots → niche developer assistants (Shopify, Unreal Engine, etc.).

The chart is less about what people type today and more about what they’ll pay for tomorrow.

Closing Thoughts

Although these patterns reflect ChatGPT usage specifically — and should be read as directional, not definitive of all AI adoption globally — OpenAI’s release gives us a rare mirror into adoption at scale. The real headline isn’t that AI is everywhere, but that it’s concentrating in very human domains: learning, writing, and everyday guidance.

One way to read this: don’t chase every possible use case. Follow where the pull already exists — and ask how to build trust, context, and scale into those wedges.

Another way: today’s 2% niches could be tomorrow’s 20% markets once product fit clicks.

Do you double down on today’s pull — or bet on tomorrow’s niches?

Until next time,
Stay adaptive. Stay strategic.
And keep exploring the frontier of AI.

Fabio Lopes
XcessAI

P.S.: Sharing is caring - pass this knowledge on to a friend or colleague. Let’s build a community of AI aficionados at www.xcessai.com.

Read our previous episodes online!