Dr Hernani Costa

Posted on Feb 28 • Originally published at radar.firstaimovers.com

700M ChatGPT Users Expose Your AI Strategy Gap

#ai #automation #business #strategy

Your AI strategy is probably built on the wrong 23% of usage data.

OpenAI's landmark study of 700 million weekly active users reveals a massive disconnect: while most European SMEs invest in coding assistance and content generation, 77% of actual ChatGPT usage concentrates in decision-making, communication refinement, and information discovery. This gap between where companies deploy AI and where users extract value represents both a strategic liability and an immediate optimization opportunity.

What 700 Million ChatGPT Users Reveal About Your AI Strategy

OpenAI's Landmark Usage Study Exposes the Gap Between How People Actually Use AI and How Most Businesses Deploy It

77% of ChatGPT Usage Concentrates in Three Business-Critical Categories

The latest ChatGPT usage data, from the largest study of consumer AI usage ever conducted, should change how every European SME thinks about AI adoption. The NBER working paper, produced with Harvard economist David Deming, analyzed 1.5 million conversations from ChatGPT's 700 million weekly active users.

The concentration is striking. Three categories dominate all consumer AI interactions:

Category	Share of Usage	Subcategories
Practical Guidance	28.3%	Tutoring/teaching (10.2%), How-to advice (8.5%), Health/fitness (5.7%), Creative ideation (3.9%)
Writing	28.1%	Editing text (10.6%), Personal communication (8.0%), Translation (4.5%), Summarization (3.6%), Fiction (1.4%)
Seeking Information	21.3%	Specific info (18.3%), Product research (2.1%), Recipes (0.9%)

The remaining 23% splits across technical help (7.6%), multimedia (5.9%), self-expression (4.3%), and other uses (4.5%).

In my experience working with European SMEs, most companies build their AI strategy around coding assistance or content generation. This data shows their customers and employees are doing something entirely different. They ask AI for guidance, not output.

Information Seeking Nearly Doubled in 12 Months, Signaling a Search Revolution

The most consequential trend in the OpenAI study is the explosive growth in information seeking. This category grew from 14% of all ChatGPT conversations in July 2024 to 24.4% by mid-2025. That represents a 71% increase in share within a single year.

The researchers themselves noted that information seeking "appears to be a very close substitute for web search."

For business leaders, this shift carries immediate strategic implications. Your customers are not just Googling your product category anymore. They are asking ChatGPT, "What should I look for when choosing a health plan?" or "What are the best options for X in the Netherlands?" These conversational queries bypass traditional search engine results entirely.

Answer Engine Optimization Becomes a Business Imperative

Traditional SEO focused on ranking for keywords. The information-seeking surge demands a parallel strategy: making your business the answer that AI systems cite. Companies that structure their expertise as clear, authoritative, citable content will capture this growing channel. Companies that rely exclusively on Google rankings will watch a significant portion of their discovery pipeline evaporate.

Writing Declined from 36% to 24% While Editing Surged

Writing was once ChatGPT's dominant use case. It is not anymore. The writing category dropped from 36% of all usage in July 2024 to 24% a year later. But the real insight sits inside the numbers.

Two-thirds of all writing-related messages involve modifying existing text, not generating new content from scratch. The breakdown tells the story: editing and critiquing provided text leads at 10.6%, followed by personal communication at 8.0%, and translation at 4.5%. Pure creative generation, the use case that dominates AI marketing headlines, accounts for just 1.4%.

This pattern matches what I see across the companies I advise. The teams generating the most value from AI are not asking it to write blog posts from nothing. They hand it a rough draft and ask for improvements. They paste a client email and ask for a more professional version. They take meeting notes and ask for a structured summary.

AI Literacy Training Must Reflect Actual Usage Patterns

If your workforce AI readiness program teaches employees to prompt AI for original content generation, you are training them for the minority use case. Effective AI literacy training focuses on the skills that match real behavior: editing, refining, translating, and restructuring existing work product. The 10.6% editing category alone represents billions of daily interactions where employees are already augmenting their output.

70% of AI Usage Is Personal, Not Professional, and That Changes Workforce Strategy

The OpenAI study found that approximately 70% of ChatGPT conversations are not work-related. Non-work messages surged from 53% in mid-2024 to 73% by June 2025. People use AI to navigate health decisions, plan meals, tutor their children, and manage personal communication.

This data point matters for workforce strategy in two ways.

First, your employees already use AI fluently in their personal lives. The adoption barrier is not technology literacy. It is organizational permission and workflow integration. Companies that remove friction between personal AI comfort and professional AI application gain immediate productivity advantages.

Second, your customers are forming AI habits outside of work that shape their expectations inside of it. A customer who asks ChatGPT to compare insurance plans at home will expect your sales process to match that conversational, advisory experience. The 49% of messages classified as "Asking," where users seek advice rather than task completion, confirms that people value AI most as a decision advisor.

Coding Accounts for Just 4.2% of Consumer AI Usage

The technology industry's narrative around AI focuses heavily on code generation. Developer tools, copilots, and AI coding assistants dominate conference stages and investment announcements. The actual consumer data tells a different story.

Computer programming represents 4.2% of all ChatGPT messages. Combined with mathematical calculation (3.0%) and data analysis (0.4%), the entire Technical Help category reaches just 7.6%. This category also shrank significantly, declining from 12% in July 2024 to around 5% a year later.

For European SME leaders evaluating AI investments, this data provides a critical calibration point. The AI tools and strategies that generate the most organizational value will not center on code. They will center on the categories where 77% of usage already lives: helping people make better decisions, communicate more effectively, and find information faster.

AI Investment Strategy Must Follow Usage Data, Not Hype Cycles

I constantly see companies burn through AI budgets chasing the latest technical capability. The ones that succeed ask a simpler question: "Where are people already getting value from AI, and how do we channel that into our business processes?" A comprehensive AI Readiness Assessment can provide the clearest answer.

Nearly Half of ChatGPT Users Are Under 26, Reshaping Talent Strategy

The demographic data in the study carries long-term strategic implications. Nearly half (46%) of all messages sent by adult ChatGPT users come from people aged 18 to 25. This is the first generation that will enter the workforce with no memory of professional life without an on-demand AI advisor.

The gender gap has also closed. Early ChatGPT adoption was roughly 80% male. By mid-2025, users with typically feminine names represented 52% of the active base, reflecting the general adult population.

For talent strategy, these numbers signal a fundamental shift. Your next wave of hires will arrive expecting AI-augmented workflows as a baseline, not a perk. Companies that lack mature AI integration will face a competitive disadvantage in attracting talent that has spent years developing intuitive AI collaboration skills.

The geographic distribution adds another dimension. ChatGPT adoption growth in the lowest-income countries now runs at four times the rate of the highest-income countries. AI fluency is becoming a global standard, not a Western luxury.

The Asking-Doing-Expressing Framework Reveals What AI Does Best

The researchers introduced a taxonomy that classifies every message into three modes: Asking (seeking advice or information), Doing (requesting task completion), and Expressing (sharing thoughts without expecting action).

Mode	Share of All Messages	Share of Work Messages
Asking	49%	~35%
Doing	40%	56%
Expressing	11%	~9%

The "Asking" category is growing fastest and receives the highest quality ratings from users. People value ChatGPT most when it serves as an advisor, not a task executor.

For work-related messages specifically, "Doing" dominates at 56%, with writing as the primary task. But the growth trajectory belongs to "Asking," suggesting that as AI capabilities improve, the advisory function will overtake task completion as the primary professional use case.

This directly informs how European SMEs should structure their AI governance frameworks. Systems used for advice and decision support carry different risk profiles than systems used for task execution. Under the EU AI Act, an AI system that influences employment decisions through advisory recommendations requires different compliance documentation than one that formats spreadsheets.

Five Strategic Actions for European SMEs Based on the Usage Data

The OpenAI study is not academic curiosity. It provides a data-driven foundation for immediate strategic decisions. Here is what I recommend:

1. Build your answer engine presence. Information seeking doubled in 12 months. Structure your expertise as clear, authoritative content that AI systems can cite. This means declarative statements, specific data, and transparent sourcing.

2. Redesign AI literacy training around actual usage. Stop training employees to generate content from scratch. Focus on editing, refining, decision support, and information synthesis, the categories where 77% of real usage occurs.

3. Treat AI as a decision advisor, not a task robot. The "Asking" category at 49% of all messages reveals that people's primary relationship with AI is advisory. Build internal workflows that leverage AI for judgment enhancement, not just automation.

4. Prepare for AI-native talent expectations. With 46% of users under 26, your hiring pipeline expects AI-augmented work. Companies without mature AI integration will lose candidates to those who provide it.

5. Align AI governance with actual risk profiles. Advisory AI and task-execution AI require different compliance approaches under the EU AI Act. Map your AI risk assessment to the actual ways your organization uses these systems.

DEV Community