It has been for a long time.
PageRank was never designed for adversarial scenarios.
It reminds me of KPIs like lines of code - it's only useful if it cannot be manipulated.
"When a measure becomes a target, it ceases to be a good measure."
PageRank was never designed for adversarial scenarios.
It reminds me of KPIs like lines of code - it's only useful if it cannot be manipulated.
"When a measure becomes a target, it ceases to be a good measure."
That all seems pretty quaint these days, but even worms and viruses at the time were versions of "my goodness, who would ever write a script that emails itself to your address book, or copies itself across the network to all the other unsecured PCs with passwordless full hard drive access?"
The real consequence PageRank had was that it got web pages to stop linking to each other…. Google gaslighted people into removing the competition they could have had navigating from one page to another through links. (e.g. web directories)
I think the problem may more have been the lack of a sustainable revenue model. Getting volunteers to curate the directory is particularly destructive because the people who most want to volunteer either (1) want to promote something or (2) want to get paid so somebody else can promote something.
That's what search indexes are. Automatically curated directories.
It's just that the UI to the index is natural language queries instead of clicks through a graph, because the latter isn't scalable to large corpuses.
There are certain sites which have a landing page for every topic in some set of topics, you could add a lot of links quickly if you built rules for importing links from particular sites. Adding 10 sites a day with 1000 links each would be very possible, in 100 days you could build out a million links.
Google was caught flat footed and wound up buying and killing Freebase in order to catch up. They lied and said they rejected "semantic web" approaches despite hiring one of the leaders of the Cyc project as their head of research.
Still there is a difference between exposing that kind of database through a full text index vs exposing it through a browsing interface.
When there’s so much money at stake, people will go to great lengths to reverse-engineer the things that put them higher in the page, no matter how you try to hide the levers that make the rankings work.
That just turns it into a statistical problem.
I don't think you can escape the problem of a search engine being used as an oracle.
Instead, you want to throw queries and user behavior into a blackbox algo and have it tell you a result then give it feedback on whether the result was good (did the user come back and ask the same question? did they click the top result and leave or did they have to come back and click through many more?). I think this is kind of how google works now, though results are frequently meh. Millions of backlinks will get you noticed but your ranking will just keep dropping if users don't appear to find your content useful (e.g. they hit the back button a lot and keep searching).
What stops me from writing a bot to simulate that behaviour for my website (or websites of my competitors) ?
One form of personalization that they applied early on is move results you click on a lot up in your results you can’t trust the serps you see at all.
They also inject a lot of arbitrariness and randomness to make it impossible to make small changes to your site to incrementally improve it the way you would improve an advertising campaign. One reason the web seems so frozen is that if you have a successful site and make major changes to the layout, titles, link structure, etc. you will possibly trigger the ‘chaos monkey’ and wreck your ranking and it may never recover.