How to influence Copilot?

[Mark_A]

Google's Copilot is muddling our company with another company with a similar name.

This is causing phone calls to us asking for the other companies products and probably vice versa.

The domain names of the two companies are different but Copilot seems to be ignoring that.

How to change this, any ideas?

[Mark_A]

I just engaged with Copilot and told it where it was making mistakes, it seems to have accepted this and promises to do better next time! :)

[engine]

>promises to do better next time!

LOL. if only.

On a more serious note, i'd probably look more closely at the errors and try to see if there are some obvious changes that could be made to emphasise the "strength" and "Validity" of the company name.

Here's an obvious example, "Widget maker" to "Green Widget maker in London"

Take a look at the other company and see what similarities exist. If you're getting calls, it's quite likely the other company is getting them, too. Most likely the information has been crawled and extracted from their website, along with other sources.

I hope it gives you some ideas.

[Mark_A]

Will have a play and a think. At the very least it shows a lack of precision in Copilot, but also there are issues that there are some similarities between the two companies at least at a trivial level.

[engine]

That's one of the challenges to influence it, is how it interpreted the data, which could come from many sources out of your control.

It's not just copilot, but other LLMs which shortcut the Web pages from an originating site.

[Mark_A]

Hmm, of course copilot is Bing not Google ..

I knew that ..

:)

[engine]

>of course copilot is Bing not Google

LOL, yep, easy mistook to make, but the approach should remain the same. The data is scraped from many sources, and one is the website you control so you might as well do as much as you can.

FYI, i'm also seeing some slightly stale info on AI answers.

[Mark_A]

Of course my pointing out details relating to two domains and the contents of their sites couldn't be learn't as in retained by copilot.

Wait a mo, why not? All copilot is doing is regurgitating information about domains which it has obtained by various means, so why not retain mine?

Associate it with the domain and retain it ..

[engine]

>why not retain mine?

It needs to be crawled (scraped) in the first instance.
Check your site logs for GPTBot [platform.openai.com...]