What people actually use AI for

The Anthropic Economic Index classifies millions of anonymized Claude conversations by topic. The pattern is clear: people use AI to create content, learn, build software, and research. For a B2B SaaS vendor, every one of those top topics is a moment where your category can come up.

Add it up: content creation (22.7%), research and intelligence (10.9%), and knowledge retrieval and enterprise search (3.6%) together account for 37.3% of all AI conversations. More than a third of AI usage is people creating information or finding it. The two most common individual work tasks in the entire dataset are both search tasks: searching databases and electronic sources (4.95% of sampled conversations) and answering reference questions from online sources (3.74%).

People are not just drafting emails with AI. They are asking it questions and acting on the answers. That behavior is exactly what Lureon’s answer engine optimization service exists to capture.

How often does AI recommend products and services?

This is the heart of the report. Six of the 50 most common work tasks matched in AI conversations are recommendation or purchase-advice tasks. Combined, they represent roughly 6.4% of all sampled conversations worldwide. Every one of them is a moment where an AI assistant names specific vendors.

There is one more number worth sitting with. Of all the artifact types AI conversations produce, "advice or a recommendation" accounts for 10.7%, and "explanation or answer" for another 16.7%. Over a quarter of AI output is the assistant telling someone what to think or what to pick. This is the layer Lureon works in: making sure that when the pick happens in your category, it includes you.

Which countries use AI the most per capita?

The Anthropic Usage Index measures a country's share of Claude usage divided by its share of working-age population. A score of 2.0 means twice as much usage as population alone would predict. The heaviest adopters are the exact markets most B2B SaaS companies sell into.

The United Kingdom sits at 3.35×, Germany at 2.40×, Japan at 1.91×. In other words: the top 12 AI-adopting countries per capita overlap almost perfectly with the highest-ACV B2B SaaS markets in the world. If your pipeline comes from Australia, North America, or Western Europe, your buyers are already living inside these tools at 4 to 6 times the global baseline. These twelve markets are where Lureon focuses its AI visibility work for B2B SaaS clients, because this is where an AI recommendation converts to pipeline fastest.

Do people work with AI or delegate to it?

The dataset splits every conversation into augmentation (the person stays actively involved) and automation (the person directs the AI to complete the task). Worldwide, it is nearly an even split.

One more classifier estimate from the dataset: tasks that would take a person roughly 5 hours working alone ran to roughly 40 minutes of conversation with AI. Whatever the exact multiple, buyers now compress an afternoon of vendor research into a coffee break. The window in which your content can influence a deal has shrunk to the length of one AI answer.

What independent research says about AI-driven software buying

The Anthropic Economic Index measures the demand side from actual AI conversations. Independent industry research, from buyer surveys rather than usage data, lands on the same conclusion: buyers have moved into AI assistants, and most vendors have not followed them.

Read those four numbers together and the recommendation gap stops being a theory. Half of buyers open an AI assistant before they open Google, yet even brands that dominate traditional rankings rarely surface in AI answers. TrustRadius's 2026 Buying Disconnect report adds the nuance that matters most: AI has changed how buyers research, but not what they trust. Assistants lean on the same reviews, comparisons, and expert content buyers always trusted. Winning the recommendation means being present in both places, which is exactly the work Lureon does.

Why most B2B SaaS brands miss the AI recommendation

The demand side is measured above. The supply side is where the gap opens: most SaaS websites are still optimized for a ten-blue-links world, while the recommendation now happens inside ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Overviews. Four failure patterns show up again and again in Lureon audits.

Rankings don't pay you. Revenue does. The question is no longer "do we rank for our category keyword" but "when a buyer in Sydney or Zurich asks an AI to evaluate tools like ours, are we one of the names it says out loud". Lureon's work is led by a senior operator with 12+ years in search, everyone else on the team has 6+ years, no juniors, no outsourcing. We measure clients on one thing: showing up in the recommendation.

Frequently asked questions