Microsoft’s Copilot rarely reproduces even full sentences from news articles and books, let alone substantive chunks that could substitute for the original, the company says in new legal filings as it fights copyright claims from publishers including The New York Times and book authors. As part of the lawsuit’s discovery, Microsoft provided 8.2 million Copilot chat logs to an expert hired by news publishers. The logs, it claims, were specifically chosen “because they hit on keywords implicating use of News Plaintiffs’ websites, and therefore the most likely to contain News Plaintiffs’ works.” It says the resulting analysis shows that 59,545 — or about 0.73% — of the logs contained excerpts from the New York Times. That’s roughly equivalent to a single article per day being pulled from Copilot’s responses over the sampled period. So, if you thought Copilot was drowning in NYT content, you might want to grab a snorkel and a life vest—because it turns out the water is shallower than you’d expect.

But let’s not get too cocky. The 0.73% figure is based on logs that were handpicked for their keyword relevance to NYT sites. It’s like saying you’ve tasted every flavor of ice cream in the world because you tried the ones labeled “chocolate” and “vanilla” at an ice cream shop. Sure, those are popular flavors, but what about the obscure ones like “dragon fruit basil” or “black garlic”? Similarly, Copilot might be missing out on less obvious NYT gems that weren’t captured by the chosen logs. And let’s face it, a single NYT article per day doesn’t exactly scream “viral sensation.” It’s more like the daily horoscope—nice to read, but hardly groundbreaking.

Moreover, the analysis only accounts for excerpts, not full-fledged reproductions. So while a sentence or two from the NYT might pop up in a Copilot response, it might not capture the full narrative or depth of the original article. It’s like summarizing a novel in a single sentence—possible, but does it do justice to the plot twists, character development, and emotional resonance? Not always. Plus, the logs were cherry-picked, which means they might skew the percentages in favor of NYT content. It’s like choosing only the best photos from a vacation album to impress friends—everyone loves the highlights, but what about the less photogenic moments?

Lastly, the 59,545 excerpts from NYT out of 8.2 million logs equate to about 0.73%. That’s roughly one article per day, which could be considered a light bite compared to Copilot’s daily intake of user queries. But here’s the kicker: if the logs were indeed selected for NYT keywords, maybe the real question is whether Copilot is just nibbling at the surface or truly digesting the substance of news articles. So, while Microsoft can proudly proclaim that virtually nobody was grabbing NYT articles through its chatbot, the evidence might just be a smattering of NYT goodness sprinkled across Copilot’s vast culinary landscape.

In essence, Copilot’s relationship with the New York Times is like a brief, flirtatious encounter rather than a deep, committed romance. The numbers suggest that NYT articles are a tasty side dish rather than the main course in Copilot’s repertoire, leaving room for speculation on whether the chatbot will eventually savor the full richness of news content or stick to the familiar flavors it’s already enjoying.