Hacker News Reader

Comment by rglover | Hacker News Reader

How Google can fix their search: embedded ranking.

Give me an embeddable iframe that I can optionally add to my site which allows visitors to give feedback (think a floating Reddit upvote button). Require that button's access to require an authenticated user via Google.

Ranking in the algorithm is weighted such that the organic user votes are the heaviest, and content length, keywords, etc., are the lowest.

Remove all the AI crap (or tweak it so I can chat with a bot to improve my search, but it's not Foie Gras'd down my throat). Make ads a free vs paid experience (want to avoid ad results, pay Google $5/mo for a clean result set).

This would make the only way to "game" SEO authentic, quality content. In essence, it's taking the current hack (appending "Reddit" to the end of a search query) and building it into the core experience.

View on HN

99% of websites won't implement, promote, or get engagement with this universal button.

And so then the way to rank is to get in a group chat with all your SEO buddies and trade ~~backlinks~~ upvotes with each other. Or make bot accounts that appear real.

Even if this idea worked, the idea of giving Google even more consolidated power/knowledge over the open web is horrifying to me.

So those sites will be ranked accordingly.

Upvote farming could be mitigated by vote frequency. That sort of behavior would be obvious in the data as you'd see significant voting activity over some time range (which wouldn't look like organic search behavior).

You can't outright avoid the consolidation of power with search engines, unless (and I think this would be smart for someone to do), you create a decentralized version of what I described that's part of an alternative search engine.

Google has been playing this game for 20+ years and failed hard the last 10+. You think you could easily solve it?

I'm not saying they're applying the best of ideas and doing it well, but I'd wager none with good experience in this field would bet it would be that easy. I think it's more akin to hackers and defenders, each surpassing the other one in a never ending race.

> Google has been playing this game for 20+ years and failed hard the last 10+. You think you could easily solve it?

Yes. Google hasn't really produced anything of substance post-2010. They fell into the typical corporate bureaucracy trap. The rejection of ideas like mine are likely even more common internally in Google. In an org of that size/stature, people jockey for position, they don't create (and that's not an opinion, it's objective reality).

O.G. Goolge would try anything (I remember they offered radio station automation software briefly).

Could I be wrong? Sure. But if public opinion of my core product is waning massively, I'd rather trust the mad scientist than the Wharton dork who's focused on his L-status. Worst case scenario the failure is a rounding error on my balance sheet.

Maybe, although I think your understating the complexity that filtering through data at Google scale can involve. When you have enough users in a system many many 'normal' people will look and behave like bots/fraud, things that should be obvious are often not.

I wonder if Stumbleupon could have achieved something similar to what you're proposing if they had tried to build a search engine.

Organic ratings are only valuable to the extent that they are not manipulated. At Google's scale, you attract click farms: https://www.fraud0.com/wp-content/uploads/2023/02/Smartphone...

This could potentially be mitigated by observing vote activity from IP ranges and frequency. Throttle vote activity for a range based on frequency over X window of time. Banning accounts could help, too, as well as more stringent account verification (e.g., want to enable voting on this account, give us a phone number or other form of 2FA).

It could get murky, but you could even have votes weighted by an account score (not quite sure how that would take shape).

They’ll give you a phone number then and enter totp and solve captcha. And click as organically and humanly as google is concerned, with sleep schedules and days off, with click hesitation and cursor movement profiles. It’s not that big of a deal, all instruments are available.

Spammers are really good at using other people's computers and accounts.

If anybody has/can solve that problem, it's Google. I think their deficit is vision, not implementation.

Google has far better signals than that. Organic cross traffic, time people spending on the page etc. serves the same purpose and that info is available for nearly free. Adding voting mechanism makes the whole business too expensive. (Imagine maintaining a complex, frequently mutable table column for every page on internet).

No different than their analytics approach. The only difference is manual vs. automatic updates (and those updates can be queued as the computed result doesn't need to update immediately).

On other words, now that the big players have taken most of the easy fruits, build the long-term long trail end of useful constructions where there isn’t much money to work with.

I don't think you can do this as part of the page. You can make the iframe almost invisible and place it directly under the user's cursor at all times. And the iframe would be none the wiser: as soon as the user clicks anything, they "vote".

Ultimately this needs to be part of the browser chrome, since you can't trust the site.

I think you could solve this by using some trickery to know the position of the element in the window. Maybe an iframe isn't exactly it but something. I think you're right, something in the browser would be best (or at least an extension).

Google weirdly does that already, in a way. It tracks when you come back quickly from a page to the search results and interpret this as a negative feedback ( not a positive one ).

> Require that button's access to require an authenticated user via Google.

Wouldn't that mean that tons of people won't be able to "vote"?

Hypothetically. But that isn't necessarily a bad thing. I'd rather have things rank based on a limited sample size than none at all.

Reply on news.ycombinator.com