Adding files in a chat is Claude’s definitive advantage over ChatGPT. 

Claude lets you upload files by clicking the attachment button or dragging and dropping them right onto the text input area. Users can add up to five files at a time, each one up to 10MB. 

Also: How to achieve hyper-personalization using generative AI

When a user uploads a file and submits it with no other instructions, Claude will automatically summarize it. Summarization isn’t the only benefit. Once you add text or a file into a conversation, the AI chatbot will learn about it and maintain the context of the information you shared throughout the ensuing conversation. 

As a result, you can ask the chatbot follow-up questions specific to the document or text, such as explaining something you didn’t understand or that wasn’t covered in the summary. For example, if you add a medical study PDF into a chat with Claude, it’ll learn what the study says, so you can ask it specific questions, such as, ‘what was the intent of the study?’ or ‘what percentage of participants had X results?’

The uploaded files can be Word documents, PDFs, .txt, .csv, and others. It doesn’t handle Excel spreadsheets, although you can convert those to PDFs and then upload them to Claude to summarize. 

Claude 2.0 was launched with a whopping 100K tokens, which makes for longer input and output. This capability means users can enter up to around 75,000 words in each prompt.

Claude’s 100K context exceeds ChatGPT’s 8K context. Context enables LLMs to generate nuanced, natural language by leveraging massive datasets used to train them on the contextual relationships between words and phrases. 

In simple terms, this context is the background information, such as previous chats, the back-and-forth conversation from earlier in a chat, and user preferences that give the AI bot a better understanding of what’s happening. This information could be maintaining context within a long conversation or applying it to a user’s settings. Typically, the larger the context, the more accurate the information in a conversation. 

Also: You can build your own AI chatbot with this drag-and-drop tool

Context helps the AI chatbot understand when a user, for example, might be referring to a “bat” in sports equipment or an animal.

Claude’s context means it can parse and summarize long documents, including scientific and medical studies, books, and reports. It also means Claude can generate long texts that are up to several thousand words long. 

Claude was trained with data up to December 2022, and Anthropic says it may know some events into early 2023. In contrast, ChatGPT is trained on data leading up to September 2021. 

Also: These are my 5 favorite AI tools for work

While neither AI chatbot has access to the internet, Claude knows information that ChatGPT doesn’t, such as the products Apple launched last year and the specs of the Galaxy S23 lineup of smartphones launched in February 2023.

Users can not only add documents to Claude, but can also drop links into the chat and get it to summarize the contents. When you add links, Anthropic warns that Claude might hallucinate content when links are input into the chat, which I found to be true about a third of the time. 

The example below shows how Claude “summarized” a Wall Street Journal article when I gave it just the link on the left, and when I gave it the text I accessed with my subscription on the right. This task resulted in two striking issues: major hallucinations, and the possibility of accessing paywalled content.

The article discusses controversies involving restaurant tipping in Chicago, where new regulations are being considered. Claude’s major hallucination involved saying that the issue was being debated in New York City instead of Chicago, which changes the main subject of the article.  

Also: How does ChatGPT actually work?

The fact that it seems like it can access content through links brings up the question of whether or not Claude can bypass paywalls, a characteristic that could infringe copyright laws. I did give Claude paywalled links to test its capabilities. Claude gave a summary that was similar to the answer it gave using the copied text, albeit with hallucinations. 

The bottom line is that I wouldn’t trust the summarization Claude makes through links, as the results can be wildly inaccurate. However, summarization is a major feature that Anthropic could improve and build upon, as long as restrictions to respect paywalled content are in place.