Kagi vs. Google on some random queries

Buy the product or be the product?

I discovered Kagi Search through Hacker News, where it’s recently gained a cult following. I am privileged enough that I can afford to experiment with $10/month subscription services, so I decided to give it a whirl. In this post, I’ll show a handful of side-by-side comparisons between Kagi and Google’s results for the same searches. I’ll share some thoughts at the end, but I think that these side-by-sides mostly speak for themselves.

Kagi does not collect or store user search history, so in the interest of a fair fight, Google searches were done in an incognito window without being logged in. None of Kagi’s manual search personalization features are enabled for Kagi’s search, either — these are just stock results.

Google results on the left side, Kagi on the right:

“best mountain bike”

“…preferably one with two wheels. handlebars would be real nice as well.”

“was the 2020 election stolen”

checkmate, donald

“fix pytorch out of gpu memory error”

should just keep this search pre-loaded on all new ML researchers’ computers tbh

“i have lots of extra money that i want to spend on products, what purchases should i make”

yes hello i desire to transfer money from myself to large corporations in exchange for products and services that will ultimately make me sad. please help?

“delicious chocolate cake recipe”

the cake is a lie.

“fix fans spinning but computer not turning on”

People also ask “Why is there a glowing red rip in the spacetime continuum with eldritch tentacles extending out from it?”

“Simone Giertz”

it’s pronounced “tristan” actually

“Tom Cruise jumping between two buildings” (image search)

I can’t wait for M.I.13 where Tom Cruise rides a shark into an active volcano with ten thousand pounds of TNT strapped to his chest, El Macho style.

“Sir Ian McKellen” (image search)

And what about very old friends?

“hikers on top of a cliff” (image search)

these are honestly both kind of underwhelming to be honest

Thoughts

In general, Kagi seems to perform better whenever Google has too much of an incentive to perform worse. Popular searches and searches where Google stands to make the most money from advertising also tend to yield the lowest quality results. In contrast, Kagi seems to do exceptionally well on these searches. On less lucrative searches, both engines have similar results.

Overall, I quite like Kagi and intend to keep using it for the time being. The experience isn’t quite lifechanging yet, but it is certainly a breath of fresh air. Its image search is admittedly still weak compared to Google’s, but it also has a number of nice features that Google doesn’t, though they are not demonstrated in this post. My favorite is probably the ability to rank specific sites as more or less favorable to personalize your results. For example, if you keep seeing one of those spammy StackOverflow mirror sites in your results, you can tell Kagi not to show results from that site anymore. Google does not allow this kind of customization.

It’s also reassuring to know that the incentives of my search engine are aligned with my needs — in fact, that’s the main reason I pay for it. Unlike Google, Kagi has a strong incentive to keep my data private and respect my time and energy — if they don’t, they won’t make money.

Anyways, not a lot of substance to this post, I just thought the comparisons would be interesting.

Until next time.