Dani AI Discussion

Also, no offense to @rproffitt, but I am going to favor Google Analytics graphs like this and usability studies with Microsoft Clarity any day of the week over someone being loud. I don't agree with downvoting people simply because Dani AI was added to their thread years later, or complaining over and over again that Dani AI needs an off switch, yet giving absolutely no valid explanation why they believe it's harmful. I don't respond favorably to loud people. I do, however, respond to people with compelling arguments and data to back them up.

@Dani - I know you understand brain fog. I've had about three hours sleep since 8:00 yesterday morning so I thought @rproffitt was downvoting the original post rather than tha AI summary. Apologies. I'm still running on empty so I'll just stop there.

Hmmm. Why is the @Dani a link in my last post and not @rproffitt? Or in this post?

Feel better @ReverendJim! Only members who have posted in the thread have their @mentions link to a popup of their posts in the thread.

Here's my take.

1. Have a way to vote on the AI content. Or at least report it if we disagree with the content. As is it appears to be unregulated.
2. The AI reply was IMO awful.

I was somewhat bemused by how it managed to spin out an entire page full of text from just 4 words - words which were likely only there to meet the minimum post length, since the question was in the title.

With extrapolation like that, I feel we're close to a TPV moment.

The thing that pissed me off the most about it was the "As OP hinted" opening. The OP didn't hint a damn thing, it's just a garbage keyword post that should have sank without trace.

The 4-word post should have been deleted, not fed through a word factory.

This just in. If you dare call out AI or check facts, you may put your visit to this country at risk.
Here I already was censored for daring to write about what's going on in this country.

• The US will now deny visas to fact checkers

The 4-word post should have been deleted, not fed through a word factory.

It was fed into Dani AI because it had received an upvote. For lack of a realistic way to self-evaluate 250,000 threads, my algorithm assumed that being upvoted meant that it most likely had enough merit of some sort to be shown as a recommended topic and fed to Google. Dani AI at least gave it some substance, which IMHO is infinitely better than where it started.

Have a way to vote on the AI content.

I have been mulling over that idea for awhile. I’ve ultimately decided against it for now. It’s not in my short term list of things to do because it would take a significant amount of coding effort, and Dani AI is not seen by logged in members (who have the ability to vote) very often. After all, Dani AI is generally not made for threads with replies newer than 2 years old, and most logged in members tend to hang around the more recent topics.

A very, very small fraction of posts that members come across ever get voted on. Couple that with how infrequently Dani AI topics will be seen by potential voters, I cannot imagine that Dani AI topics will be voted on much at all.

It was fed into Dani AI because it had received an upvote.

Let me rephrase that. It was fed into Dani AI because it had received more upvotes than downvotes. A thread with 5 upvotes and 10 downvotes would not be fed into Dani AI.

The thing that pissed me off the most about it was the "As OP hinted" opening. The OP didn't hint a damn thing, it's just a garbage keyword post that should have sank without trace.

I’m interpreting this sentence to mean, how dare Dani AI take completely horrid content posted by humans, and then left by humans to rot, and at least try to improve it.

Why not try to use automation to at least attempt to salvage content that was created by humans, left to rot by other humans, but upvoted (and deemed to have merit) by yet other humans?

Story time: Went to see the Slate EV Truck in San Diego with others in the area. We turned on Grok in the car and hilarity. Grok isn't ready. The AI I see here is isn't ready. Unless you measure readiness in non-human metrics.

Again there is no way to report AI slop auto-generated here so you are left to downvote what you can.

It was fed into Dani AI because it had received an upvote.

Can you tell who upvoted?

Was it some similarly low post count drive-by account who hasn't been back in as long as the OP?

"Anonymous" +/- are so cheap and easy to apply, it hardly counts as endorsement. When compared to say actually pressing the "comment" button, typing in some actual words and pressing +/-.

Also, how about only summarising actual conversations - you know, two or more people posting two or more messages. Those might actually be worth a summary. Single post drivel with zero engagement, no matter how upvoted, should just stand by itself with whatever quality (or lack) it inherently has.

After all, Dani AI is generally not made for threads with replies newer than 2 years old, and most logged in members tend to hang around the more recent topics.

Which is all well and good until some goof-ball shows up just to dribble some keyword spam on it. Then we all see it and start another round of conversation.

my algorithm assumed that being upvoted meant that it most likely had enough merit of some sort

For fun, I checked some of my really old replies from decades past.
https://www.daniweb.com/programming/software-development/threads/41868/anyone-could-help-me
https://www.daniweb.com/programming/software-development/threads/41829/regarding-dynamic-memory-allocation
https://www.daniweb.com/programming/software-development/threads/41819/segmentation-fault
Not a single vote to be seen, what was your metric here?

"Anonymous" +/- are so cheap and easy to apply, it hardly counts as endorsement.

Too easy. As I recall many years ago one member got pissed at another and downvoted hundreds of his posts to trash his reputation. IMO anonymous downvotes are too easy to apply. Dani might have put some type of limiter in place after that.

I truly don't get all the Dani AI hate. What am I missing? (Genuine question. No attitude. I must be missing something.)

Perhaps an analogy is in order. Hour-long TV shows used to be roughly 55 minutes of show and 5 minutes of ads. It is now 40-41 minutes of show and 19-20 minutes of ads. And during the actual show we are constantly annoyed with pop-up and slide-in ads that do nothing but take us out of the show. Perhaps the AI hate comes from people who see the same trend; more and more content is being generated by AI. I know that in some cases the AI consists of a generated summary of existing content, but we are seeing more and more original content being posted that is actually AI generated. At what point will we start seeing AI generated summaries of AI generated content?

I guess I just don't find them that intrusive. I can always scroll on by. I've learned a thing or two from them. If it brings more eyes to DaniWeb, that's a plus for all of us - potential for more active members and more likelihood DaniWeb sticks around? I was skeptical at first, but I don't see the problem. It's just in one scrollable area.

Would I like options? 1) the requested on/off switch; 2) ability to initiate an AI review. Sure, because others take issue with it and it would be great if everyone could be accommodated. I think it would a good programming challenge for Dani - if she's up to it. It'll keep things fresh.

Can you tell who upvoted?

It's meant to be anonymous. Nevertheless, I did a database query and saw it was a member with 0 posts.

"Anonymous" +/- are so cheap and easy to apply

Then why do so few people vote?? That's why when we do see a vote, we take it seriously.

Not a single vote to be seen, what was your metric here?

OK, bear with me for a little SEO lesson.

DaniWeb has nearly half a million forum threads. However, back in 2011, the Google Panda algorithm was released, and one of the things that it did was assign a site-wide penalty to any website in which the majority of pages are considered low quality content, or pages without substance (known as "thin content"). However, if you add the noindex meta tag to the webpage, and tell Google not to index the page, then that page does not count against your ratio of quality pages to thin content pages. Therefore, and I am super, super oversimplifying things here, it's often in a website's best interest to noindex the pages that you want to tell Google are not worth adding to Google's index. (Google still crawls those pages, however. It's a whole thing, but I digress as it's irrelevant to my storytime.)

Anyways, out of the nearly 500,000 forum threads, we allow Google to index 270K of them, and have added noindex meta tags to the remainder. Every website and forum has their own proprietary algorithm that they use to determine the indexability status of a forum thread, but basically we take things into consideration such as whether the question has received any votes, any replies, whether it's been marked solved, etc. So that's the way things have chugged along for the past 15 years.

Fast forward to now, and Dani AI is currently in the midst of a slow rollout to all older forum threads that we are already telling Google to index. The algorithm that determines indexability status has not changed in quite a bunch of years for us.

It seems to me that your anger about the forum thread in question is misplaced. Your grievance should really be why is it that this forum thread was allowed to exist, was not deleted by a moderator, and, worse yet, passed the DaniWeb noindexing algorithm to allow it to be fed into Google, which means that the quality of this page affects, and is likely to push down, our overall site ranking in the search engines. Given all of that, 2 months ago, this webpage was really harmful to our overall SEO efforts.

By adding Dani AI to pages like this, at worst, they are unlikely to go noticed by most members nor to make a big community impact. Basically we're just additional box of text on an old page that you think should be deleted in its entirety anyways, so it's not like it's affecting your usability of the site. At best, however, they turn Google's quality rating algorithm from seeing a pretty crappy page to instead seeing a slightly less crappy page with some additional relevant content, which is enough to bring this particular page out of "thin content" territory.

Done en masse, we are raising Google's quality ranking of our entire website by simply touching on old pages that members haven't interacted with or posted to in years. A higher site ranking means higher ranking for our new pages and all pages that have had quality content in the first place without the help of Dani AI.

And the numbers don't lie. Traffic from Google has doubled so far, and the rollout is only about 50% done. Engagement rates are the highest they've ever been. Time on site is the highest it's ever been. Bounce rate is the lowest it's ever been.

I am a trillion times with you when it comes to saying that AI generated content is often crap. But DaniWeb has always been a business for me, it's my only source of income (well, aside from my husband's salary), and it's been operating in the red for years. If I can use AI on content no one cares about to lift the content that people do care about, I'm going to do that.

I think we all want Daniweb to stick around, and that means operating in the black. If AI summaries help that then I am all in favour.

From a business perspective, I have always had three masters. I need to appease Google which is where all of our traffic comes from. I need to appease first-by visitors as well as members alike, because all members start out as first time visitors before something convinces them to stick around, and I need to appease advertisers who pay me.

Dani AI seems to satisfy Google, but even more than that, it seems to be having a huge impact on visitors wirh them spending more time on the site and engaging with the site more than ever, and more and more first-time visitors are signing up and creating their first posts. If all of that can happen just by adding a box with some automated content to old pages that most members haven’t interacted with in years, how is that a bad thing?

Perhaps the AI hate comes from people who see the same trend; more and more content is being generated by AI. I know that in some cases the AI consists of a generated summary of existing content, but we are seeing more and more original content being posted that is actually AI generated.

So is the hate generalized towards the entire current Internet ecosystem? It sounds like instead of hating Dani AI, you should be hating Google AI Overviews that take up a huge chunk of the page real estate above any of the search results. But then again, they exist because Google has found that the very vast number of searchers find them beneficial.

P.S. Am I the only one who finds the graph in the OP post nothing short of amazing?!

It's a significant increase - 2 minutes to 5? In my Google searches, the included AI is truncated with a Show More button. Partial solution? I'm not seeing any AI in my Bing.com results.

If your model is to cater to Google, then there is nothing more to discuss. There's no reason for humans to be here.

If your model is to cater to Google, then there is nothing more to discuss. There's no reason for humans to be here.

My model has not changed in 20+ years. DaniWeb started because of my love of SEO. I spent most of my career traveling the country speaking at SEO conferences. We have Digital Marketing and SEO forums because those are my passion. That's also why DaniWeb became so popular. Luckily, the way to rank well in Google is dependent on creating great content that engages people. So that is what I have always tried to do.

There's no reason for humans to be here.

From where I'm sitting, proactively making changes that increase the average amount of time that 95% of the humans on DaniWeb spend here increasing from an average of 2 minutes to 5 minutes is catering to humans. This argument is silly.

Recently the only engagement is to check if someone has an interesting question I can help with.
But the rise of the AI SEO AEO as the site's goals that is engaging.
As they say "You gotta do what you gotta do."

But the rise of the AI SEO AEO as the site's goals that is engaging.

I'm unsure what you mean by this. As mentioned, SEO has been a primary focus for 20+ years, and continues to be a primary focus. This was never a secret. As a platform creator, I see my role as both building the platform and being responsible for marketing it. Those things, coupled with selling advertising, is what my job description is. That's what separates DaniWeb from you downloading phpBB and slapping it up on a budget web server and expecting people to simply want to and be there to interact with you.

AEO has never been a focus and continues not to be. In fact, everything I have read about how to optimize for AI are things that I'm already doing for SEO as-is (e.g. structured schema markup, etc.).

TLA salad. Tomato, tomata, you don't put them into a fruit salad.