Does RankBrain make certain words poisonous to rankings?

[goodroi]

Has Google's AI learned to recognize the meaning of certain keyword phrases so you lose your rankings if they are added to your page? For example if RankBrain sees you mention a product is the "2017 version" does it derank it in favor of the page on your site that has the "2018 version" even though all the traditional ranking signals are still pointing to the product page for the 2017 version? Does adding certain keyword phrases trigger RankBrain to derank a page?

I'm seeing some weird ranking moves after adjusting the text on pages that do not make sense when you only consider traditional ranking signals. These weird ranking moves would be understandable if you assume Google's AI can recognize & understand certain keyword phrases.

What do you think?

[Brett_Tabke]

Wow - going to have the Poison keywords talk again? Seems like only yesterday we talked at length about it. I don't think it has anything to do with rb - it's been hard coded in G for a long time.
[encrypted.google.com...]

Checkout this classic one from 2000 with Oilman and WebG.... [webmasterworld.com...]

[JonathanEdmonton]

I never thought about poison keywords before. I'm glad I found this website. I'm learning a lot.

[aristotle]

A skillful writer can usually find a way to get the message across without using any "poisonous" words.

[goodroi]

I am not referring to old fashion stop words or profanity. From my experience it seems the trouble causing words & phrases are different depending on the type of site & industry.

From my perspective the simplest advice is to think about what web content Google want to serve to users and target that concept. We are well past basic keyword matching. Google's AI seems to be trying to make connections and understand relevancy. I try to build pages that will satisfy the user intent. When I satisfy the user I want to do so in a way that can be measured by Google. IMHO trying to provide the best user experience can lead some (but not most) people to accidentally generating the wrong signals.

[nomis5]

@Goodroi

Your original mention of poison words was associated with AI. This is definitely not the "the Poison keywords talk again" if I understand you correctly. How Googles AI treats poison words may well be hugely different from how it treated them before AI and is well worth discussion anew.

Personally I haven't noticed any effect mainly because my website is hardly likely to use any poison words.

However, today I published an article (gardening) which used the word "rape" to describe a particular type of vegetable - a normal description in the scope of the vegetable being discussed. It made me slightly uneasy given G's use of AI nowadays. I will keep an eye on the article and see how it progresses.

[goodroi]

Correct, I am not rehashing an old idea. I am bringing up a new possibility due to AI changing how Google ranks pages.

I wouldn't be too worried about mentioning "rape" if its relevant to your niche and the audience will interact with that terminology for example a page about "rapeseed for grazing". I would be more worried with a potential text saying the seed is out of stock or being recalled.

I've seen some ranking movement that is very hard to explain without making the assumption that Google AI is getting better at understanding the meaning of certain words & phrases ... or at least more prolific making a guess and trying to connect things.