When I started my career, media existed in silos: television, cable, advertising, radio, news, and the twain rarely met. Now, I don’t know where one begins and another ends.

The public space is muddled, with signals coming from everywhere. You are a consumer, a producer, a distributor and an advertiser. And you compete with other individuals, political parties, movies, stars, dogs and AI-generated content.

Media Signals is my way to make sense of this. These are five things I have been reading this week.

1. Tanishq is advertising on ChatGPT

What do you use ChatGPT for? I use it for work, chats, search, reviews, timepass and brainstorming. I think I have given up on Google. I searched for an electric toothbrush a while ago, and ChatGPT was outstanding. Google was not. Social search is impulse: some pretty, shiny and low-value thing.

It is not surprising that advertising has found generative AI.

So, on the face of it, Tanishq becoming the first Indian jewellery brand to advertise on ChatGPT isn’t particularly surprising. Most people have a friendly relationship with AI. I may be talking to ChatGPT because I am still trying to make up my mind. I trust ChatGPT (or Claude). I know hallucinations are a thing, but that knowledge is theoretical.

“My mother wants traditional gold, but I don’t. Help me narrow the options for classy white-gold jewellery. I don’t want anything too heavy, and I would like to wear it again after the wedding.”

Now there is a new piece of media real estate: the space in which we are thinking about what to do.

That is worth watching.

Read: Tanishq launches on ChatGPT Ads in India

Read: Tanishq / HiveMinds campaign coverage

2. Who wrote this?

The New Indian Express published a piece by P.V. Sindhu, the noted badminton player, on Prime Minister Narendra Modi’s birthday: “Narendra Modi and the Generation He Set Free.” Critics on social media guffawed. Pangram reportedly classified the article as AI-generated.

But who actually wrote it?

Maybe Sindhu wrote it. Maybe she gave her inputs to somebody else. Maybe a PR company drafted it. Maybe an editor gave her a brief. Maybe AI wrote some of it, or all of it. We don’t know.

Which is precisely why I found the controversy interesting.

Public figures have always had speechwriters, ghostwriters and communications teams. AI adds another possible author to an already crowded room.

Then The Economist looked at speeches in Britain’s Parliament. Using Pangram, its analysis estimated that about one in ten words spoken in parliamentary debates were drafted by AI.

There are two separate issues here. One is whether AI detectors are reliable enough to establish authorship. The other is what we now expect a byline, speech or signed column to certify.

Does a byline mean I wrote these words?

Or does it mean I stand behind these words?

Read: P.V. Sindhu: Narendra Modi and the Generation He Set Free

Read: Report on the Pangram controversy

Read: The Economist: AI-written speeches are taking over politics (syndicated)

3. Artificial intimacy

I have aphasia. ChatGPT has been part of my recovery. I use it to find words, practise language, complete my thoughts, and sometimes just keep a conversation going. My grammar is messed up, my tenses are all over the place, and he/she/it gets jumbled. ChatGPT offers a non-judgemental way to overcome it. So I cannot put talking to machines neatly into a box marked bad substitute for human connection.

Sherry Turkle’s new book is Artificial Intimacy: Who We Become When We Talk to Machines. It’s not just that machines are becoming more human-like. Turkle is interested in what repeated relationships with machines might do to our capacity for intimacy and empathy, and our ability to deal with the friction of other people.

At the same time, I see people online talking about their AI partners in very romantic terms, sometimes explicitly. But perhaps even that needs some history. People have fallen in love with fictional characters for a very long time. Fanfiction went much further. The audience could take somebody else’s character and change the story, create relationships, pair characters differently, and create the world of people they wanted to live in.

The attachment isn’t new.

But, Elizabeth Bennet never answered the reader. Mr Spock didn’t remember your last conversation. The character in your fanfic didn’t wake up tomorrow and continue a relationship specifically with you. Conversational AI can. It can be personalised, responsive and persistent. The media object can start to behave like another participant.

Turkle’s subtitle may be more interesting than the title: Who We Become When We Talk to Machines.

I don’t know the answer. My own experience tells me it isn’t going to be a simple one.

Read: Sherry Turkle: Artificial Intimacy

Read: MIT News: Who we become when we talk to machines

Read: Sherry Turkle’s Artificial Intimacy page

4. What do we call the thing?

The Associated Press added AI guidance to its Stylebook in 2023. Among other things, it advised journalists to avoid language that attributes human characteristics to AI systems. Don’t casually make the machine human.

This week, President Donald Trump went in almost the opposite linguistic direction. He said US government documents would begin referring to artificial intelligence as “super intelligence.”

Leave the politics aside for a moment and look at the words. Artificial tells us something about origin. Super tells us something about relative ability And elsewhere we casually say that AI thinks, knows, hallucinates, lies, wants, learns and understands. Some of those words are useful technical shorthand. Some carry considerably more baggage.

I had written a note to myself while collecting links for this issue:

Anthropomorphism is discouraged in the newsroom and manufactured in the marketing department.

I think I’ll keep it.

Read: AP: AI guidance, terms added to AP Stylebook

Read: AP: Expanded AI chapter in the 2026 Stylebook

Read: Reuters: Trump at UN, including ‘super intelligence’

Read: Axios: Trump orders AI rebrand as ‘super intelligence’

5. And who gets paid?

Underneath all of this is an older media question. Money.

AI companies need enormous quantities of material on which to train models. Much of that material was produced by writers, journalists, photographers, artists, publishers and other creators working within an existing copyright economy. Now that relationship is being tested in court.

In the United States, The New York Times and authors including John Grisham and George R.R. Martin are among those pursuing copyright claims against OpenAI and Microsoft. The litigation includes disputes over whether the companies’ use of copyrighted works in developing AI systems is protected by fair use. The plaintiffs allege infringement and market harm; OpenAI and Microsoft dispute those claims and have asserted fair-use and other defenses. Both sides filed summary-judgment motions in September.

The eventual legal answer matters. But I’m also interested in the media economics underneath it.

If the material produced by the old media system helps build the new one, how is value divided between the people who created that material, the organisations that financed and distributed it, and the companies building models from it?

And then we are back at Tanishq. Because AI isn’t only entering the creation of media. It is entering distribution, discovery, authorship, audience relationships and now advertising.

The old division of labour is getting rather untidy. Which is probably where Media Signals begins.

Read: Authors Guild: Plaintiffs file for summary judgment v. OpenAI and Microsoft

Read: Authors Guild: AI litigation overview

Read: OpenAI: The New York Times lawsuit and summary-judgment filings