Which AI Bot Reads Best? A Washington Post Showdown Reveals Surprises

In a revealing head-to-head test conducted by The Washington Post, five popular AI chatbots—ChatGPT, Claude, Copilot, Gemini, and Meta AI—were challenged to interpret and analyze four types of complex written content: a novel, medical research, legal contracts, and political speeches. The results offer valuable insight into how well these bots actually understand what they read—and whether they can be trusted as reliable reading assistants.

Across 115 questions posed to each bot, some responses were sharp, analytical, and eerily human. Others missed the mark entirely, hallucinating details or failing to grasp nuance. Each bot had strengths in different areas, but only one emerged as the overall champion: Claude.

Claude stood out by never hallucinating—an AI achievement in itself—and was the top performer in law and medical research. It was the only model to earn a perfect 10 for summarizing a scientific paper and was praised for its contract revision suggestions. ChatGPT took the lead in political analysis and literature, with particularly thoughtful responses to Trump’s speeches and emotional insights into a Civil War novel. Yet it lagged behind in legal comprehension.

Meta AI, Copilot, and Gemini frequently faltered, either by oversimplifying content or missing major points. Gemini in particular struggled with literature, offering what one expert likened to a clueless book club summary.

The takeaway? While AI tools like Claude and ChatGPT show great promise as reading aids, they’re far from flawless. Their performance varies wildly by topic, and none scored above 70%—a D+ grade in most classrooms. For now, these bots might serve as useful companions for quick comprehension or basic analysis, but they’re not ready to replace your own close reading—especially for important legal or medical documents.

📖 Read the full article by Geoffrey Fowler at The Washington Post to dive into the test results, expert commentary, and final verdict:
https://www.washingtonpost.com/technology/2025/05/30/ai-reading-comprehension-comparison-review/