Paste the link. Skip the copy-paste.
Berges AI fetches the page behind any URL you paste and reads it as part of your message. Ask for the summary you want, then ask follow-ups.
Try this promptThe usual way to summarize an article with AI is to open it, select everything, paste it into the chat, and hope the paste is not too long. On Berges AI you skip that. Paste the link and the page text is fetched and handed to the model along with your question.
It works on any pipeline and in any conversation, which makes it the easiest way to get past the message length limit: the article does not count against it, only your message does.
Same task, less slop
I'm sorry, but I can't access external links. If you paste the text of the article here, I'd be happy to summarize it for you! Please make sure to include the full content so I can provide a comprehensive overview.
Three points: 1) The regulator opened a review into the merger over pricing in regional markets. 2) Both companies say they will offer remedies rather than walk away. 3) A decision is due in the spring. Why it matters: the timeline pushes any price changes past the next contract cycle.
Summarize a link with Berges AI
Paste the URL
Put the link in your message with your question. A few links per message are fine. The page is fetched, its main text is extracted, and the model reads it before answering.
Say what you want out of it
Three bullets, one paragraph, the numbers only, the argument in one sentence. The shape you ask for is the shape you get.
Keep asking
The page stays available for the rest of the conversation, so "what did it say about pricing" or "is the source for that claim named" works without pasting the link again.
Put this in your prompt
- → The length and shape you want, before anything else.
- → Your angle: what you are reading it for, so the summary keeps what matters to you.
- → Two or three links at once if you want them compared.
- → A specific question about the page, if a summary is not really what you need.
Things people ask
Does it read the whole page?
It reads the main text up to a size cap, roughly the first couple of thousand words. For most articles that is everything. For a very long report the model sees the beginning, so ask about a section by name or paste that part.
What kinds of pages work?
Public pages with readable text: articles, blog posts, documentation, forum threads. Pages behind a login or a paywall will not load, and neither will pages that are mostly images or video.
Is this the same as the WebSearch pipeline?
No. Pasting a link works everywhere and only fetches the pages you gave it. The WebSearch pipeline lets the model search on its own and pick sources. Use a link when you know the page; use WebSearch when you do not.
Is the page content stored?
The fetched text is used for the conversation and cached briefly so follow-ups do not refetch it. Your message and the reply are stored only if you save the conversation, encrypted at rest like everything else.