Thanks a lot!
This OpenAI article worked for me. But you need to use the correct field name. Only by returning the URL in a field called exactly “url” was I able to click the link.
I tried some named like [anything]URL
and it doesn’t worked.
Thanks a lot!
This OpenAI article worked for me. But you need to use the correct field name. Only by returning the URL in a field called exactly “url” was I able to click the link.
I tried some named like [anything]URL
and it doesn’t worked.
Are you using an Action in the GPT, or an API to qualify and return the link?
Problems from the beginning of this topic until now. We must be separated it. Because there are many sources, even the browser we use has an effect. When time passes the cause of this has also changed.
As far as I’m concerned, I’m having trouble accessing it. The website started having serious problems a few days ago to the extent that GPTs that work with websites that don’t have an Acton API are not able to function properly, such as GPTs that work with price comparisons on different websites. It turned out to be impossible to provide links, and my GPTs that were originally intended for use on the OpenAI website still refused to use the OpenAI website, including the forums. By giving reasons in copyright laws and privacy of the destination website regardless of what the command to do is.
Even though it is GPTs from OpenAI’s (ChatGPT - ActionsGPT). Refuse to provide real-usable link content, including the use of direct user links. This causes the introduction of usage to new users who face problems with inconsistent communication.
The above problem was only resolved in the past few hours. I concluded that it was not a problem with user settings. Another thing I noticed was that the problem changes in severity. I suspect that OpenAI may be working on it. I have already told OpenAI that there is an internal fix. It would be great to notify users Otherwise, when a user encounters a problem and someone suggests a fix, the method won’t work for different reasons. Or it can be fixed but over time becomes unusable.
These problems have an impact on the people who implement the solutions. It causes the understanding of the problem and the solution to be disconnected. Therefore, when there is a problem in the end user level, no one will want to help. If there is a record indicating the working period At least people who have problems and people go in to help. May be able to record data and reference time It will become a library of useful solutions.
My one complaint about this place is that when people ask about the use of GPT+, they are often pushed to the FAQ or are worried about asking. I can understand this point in order to maintain the standards of the developer community. But I want to know what we are evolving to. If it is not developed to be used or extended or made available to the general society for use.
I’ve found the answer say to use an API, that is unlikely to come from a more knowledgeable user. And that makes me feel very strange. When people ask why there isn’t a bug reporting area (words from other users that include problems in this word) here and they just answer that it’s a developer area. But I think that a good developer should approach usability problems and solve them with higher knowledge than the average user to solve the problems more sustainably or at least can take problems to develop their products. It’s not like escaping the problem of limitations to use something more complex and calling yourself a developer.
Even having a bug hunter. And then you’ll say that we don’t need that space. It means that we are looking at the problems of users very wrongly. They are on a completely different level. Now it turns out that this part is like a space for smart people who have never even used ChatGPT. If the problems with general use are simple, It’s still better to answer, but because we can’t answer each other, we’ll just go to FAQ. Even people with a level of knowledge that should be able to create websites and write code, or have complex GPTs in their store, don’t know that there are always problems with new and updated features.
Same. However, links work with the ChatGPT mobile app. The issue seems isolated to the website.
I’m experiencing this bug as well – very frustrating.
I discovered that changing the custom GPT’s “Publish to” setting from “Anyone at [my org’s name]” to “Only me” fixed the non-clickable hyperlink issue.
However, this is clearly not a viable solution for the majority of people and use cases.
Edit: fixed typo
The problem is simply this. Any url that has a ? Followed by any text, a UTM value for example or any text at all is not clickable. It also doesn’t matter if you generate the url from an API or have chat append the values after the ? - in all cases the link is not clickable.
Th irony is chat will ask it you would like to add tracking to your urls, aka UTM values which in turn make it not clickable.
It feels like some primitive attempt at controlling what privacy data goes out of chat, but who know. Highly irritating though.
When you add anything AFTER the ? you will see this in the interface, again this doesn’t matter if you action serves up the URL OR if Chat appends something after the ?
You will get this:
Great point about the parameters. That’s the same thing happening to us with Web Performance Engineer GPT. Hoping for a fix soon.
Yeah, it basically makes it useless because unless you have some discrete custom URL, you are not going to be able to add any params to easily track activity…google analytics, out of the question.
Odd right?
Also to be sure I’ve tried this method and it does not work, I believe this is incorrect:
https://help.openai.com/en/articles/8838458-why-aren-t-my-gpt-links-clickable
Our links have UTM codes and are clickable. Try here: ChatGPT - Consensus
Is there something I’m missing? Does chat prevent UTM codes somehow?
Okay interesting, they do work (in your gpt), thanks for taking the time to chime in.
Not sure what to think then as clearly judging by this long thread, and my own well tested experience - this does not work for everyone.
I’m not going to speculate at this point. OpenAI care to elaborate here? This bug makes building a gpt mod or less useless if the average builder can’t use UTM’s and therefore track activity…
@christian-salem is your UTM included in your API response or is chat adding it?
The UTM code is in our API so maybe that’s the trick?
Yeah, @christian-salem that appears to be the case, I’m trying to publish a GPT that shows both use cases but struggling.
So we know that UTM’s are not the issue - can someone here let OpenAI know about this bug?
There seem to many problems affecting the custom GPTs the builder itself. I get all sorts of weird errors when I simply try to update the knowledge.
I have the same problem. A simple GPT that does some searches and then produces a bullet list with the best matches and the corresponding links.
Most of the time the links are not clickable. Sometimes some are clickable. And interestingly the same URL is clickable once and in the next chat it is not.
Anyway, as it is, it’s a waste of time to create GPTs that should also provide links as result of their work. What sense does Bing-Search integrated in GPT have, if at the end the found links cannot be shared with the user?
This is very annoying, and I don’t know if this helps to make the whole thing popular. This is driving GPT-Builders away if an alternative emerges that doesn’t have these kinds of problems.
I really hope that OpenAI addresses this issue ASAP!
Yes, agreed, this a fundamental functionality. If a gpt is summarizing some outside data and wants to send the user there, this basically doesn’t work. The worst part is the builders spend time convincing users to try it, which is no mean feat given the hefty price…then the links don’t work. Terrible.
Not solved but I found out something.
I accidentally opened another window and tried to do the same thing that I couldn’t get clickable link. Wow, now I could get clickable link. So, again, I tried open another browser window and, boom, couldn’t get the same result.
Seems unstable.
Something is definitely off. My GPT locates auto parts. So links are pretty important. My GPT was working perfect, I have quite a few hours making links work and look presentable for users. Now today, my links are not clickable. There is no way I can validate the mass of websites I am using. Best part is, sometimes they work, sometimes not so much. Would it be possible to display a disclaimer on links to circumvent liability for OpenAI?
Okay to be clear links work, it’s when you append parameters to a link using GPT to do it. From my testing all links work, in all cases EXCEPT when paramaters are added by GPT.