2024 was the year the ai browser arrived, and i used all of them so you don't have to

ai · 2024-12-20 · 8 min

in 2024 every browser added an ai button. not a feature hidden in a menu: a button on the toolbar, a sidebar that opens by default, a prompt box above the address bar. i installed the major ones, used each for a week of real work, and kept two. the rest are gone.

this is not a benchmark. it is a diary with a verdict at the end.

what i actually used

my week looked the same in every browser: reading documentation, debugging a react app, writing a few posts, answering email, and searching for the specific error message of the day. the ai features had to help with those tasks, not with a demo.

the features fell into four groups:

• summarize this page. highlight text or click a button, get a shorter version.
• ask about this page. a chat box that can see the current tab.
• write or rewrite. compose an email, shorten a paragraph, change the tone.
• find things. tab search, history search, and "where did i read that" questions.

most browsers shipped all four. the differences were in how much friction sat between me and the feature, and how much of the page left my machine.

the ones that earned a place

arc kept its place because the ai is not the point. spaces, profiles, and the command bar were already the reason i used it. arc max added page summaries and tab renaming, and both are small enough to ignore when they are wrong. the browser did not become an ai product. it stayed a browser with a useful shortcut.

edge kept its place because copilot is in the sidebar and the sidebar is one click. for reading a long page and asking "what is the actual argument here", it is the fastest path i found. the summaries are not always right, but they are easy to check against the page that is already open. that proximity matters more than raw quality.

the ones that did not

opera had aria, a chat sidebar that felt bolted on. the answers were fine. the problem was that the browser kept asking me to use it. every new tab had a prompt. every sidebar had a suggestion. a feature that needs to be advertised inside the product is a feature that has not found its place yet.

brave had leo, and leo was the most privacy-conscious of the bunch. it also felt like a separate product living in a sidebar. i used it twice and forgot it existed. that is not a failure of the model. it is a failure of the integration.

chrome had gemini, but in my region and my account it arrived late and half-shaped. the parts that worked were useful. the parts that did not were a tab that opened a chat and lost the page context. i did not keep it as a daily browser, so the ai never became part of the routine.

firefox had a few ai experiments and a lot of policy discussion. the ones i tried were add-ons, not browser features, and add-ons are easy to uninstall. that is fine. it is also not the same product bet.

what the ai actually did well

the boring tasks won. summarizing a long page before deciding whether to read it. renaming twenty tabs into something i can find later. turning a rough paragraph into an email that does not sound angry. answering "what does this error mean" without leaving the page.

the tasks that did not win were the ones where the browser had to know something the page did not contain. "what did i read last week about x" was usually wrong. "which of these three articles is more reliable" was a coin flip with confident prose. "write this for me" produced text i had to rewrite anyway.

the privacy bill

every one of these features sends page content somewhere. sometimes it sends the highlighted text. sometimes it sends the whole page. sometimes it sends the url and a summary. the settings are not always clear, and the defaults are not always the private ones.

i turned off page access for every browser i did not keep, and for the two i kept i turned it on per-site. that is more friction than the demo suggests, and it is the honest cost of the feature. if you are not willing to manage that, the feature is not free.

the verdict

ai in the browser is a feature, not a new browser. the two that earned their place did it by making the feature small, close to the page, and easy to ignore. the ones that did not were the ones that made the ai the product and the browser the wrapper.

i do not think 2024 was the year the ai browser arrived. it was the year the browser learned to summarize. that is useful. it is not the same thing.

for the search side of the same shift, see why i stopped using google. for the model side, see chatgpt after one year.