Clare Spencer
banner
clarespencer.bsky.social
Clare Spencer
@clarespencer.bsky.social
Reporter for generative-ai-newsroom.com, figuring out when (and when not) to use GenAI in news production. For off topic thoughts go to @aclarespencer on Threads.
Axios's map of which news organisations is suing and which ones are doing deals with AI companies, as of September, is incredible.

www.axios.com/newsletters/...
November 21, 2025 at 8:42 PM
Possibly exposing myself as someone who didn't know this corner of the world exists but chanced upon a book I wanted to read only to find out it's $156.

... could someone be kind and explain to me why a book is $156 and if this is normal then what world is it normal in?
November 20, 2025 at 11:17 PM
This is how the lead for the BBC’s responsible AI team has summed up the findings: www.linkedin.com/posts/radioj...
October 22, 2025 at 8:22 AM
My MVP… posting here as a screengrab. But think it might perhaps work as a Google sheet or perhaps as part of a wider public NotebookLM (perhaps with all my transcripts. What format would you find most helpful??
October 19, 2025 at 6:11 PM
Also... It is important to note that I would only ever use generative AI to illustrate a factual article with illustrations. I draw the line at anything that the audience could mistake for real. I don't even like using stock photos for this reason.
August 12, 2025 at 5:02 PM
I have just finished a first draft of a story where I have NO real life pictures. This time I thought I would create a theme - trying to emulate what it may be like to commission an illustrator for a long feature article. Here are my pictures.
August 12, 2025 at 4:49 PM
I created the image created by ChatGPT with the prompt “Please generate a picture of extracting the criteria for what makes good and bad content “trapped inside the editor’s head”.”

Again. There is room for improvement.
medium.com/generative-a...
August 12, 2025 at 4:47 PM
Second, here's a quick audit of illustrations I've used so far. This was an illustration of a generative AI pixie adding in errors reporters have to weed out. I used OpenAI.

I think it is safe to say there is room for improvement.
August 12, 2025 at 4:42 PM
2/ The site I write for, generative-ai-newsroom.com, uses Medium, which pulls in a picture from your article on our front page of our site. It also pulls in a picture for social media (see Bluesky here). So I want that to look good so people feel compelled to click on it.
August 12, 2025 at 4:39 PM
@sarahingram.bsky.social I think you may have a factual error in your story.
August 5, 2025 at 5:38 PM
The BBC Executive Sponsor of Gen AI, Rhodri Talfan Davies, earned more than £305,000 last year, according to the BBC annual report.

www.bbc.co.uk/aboutthebbc/...
July 15, 2025 at 1:51 PM
Ex-BBC R+D staffer Michael Armstrong says he finds this quote 'troubling'.
"The BBC Sounds subtitling trial had humans checking the output, it lasted for three months and most of the audience had no idea it was happening."

www.linkedin.com/posts/activi...
July 15, 2025 at 1:29 PM
I should have mentioned my sources... here is where I heard that that story was the first to use AI to help write the bullet point summary: www.linkedin.com/posts/nathal...
July 14, 2025 at 5:41 PM
At the bottom of the article is a sentence saying that BBC News used AI to help write the summary.
July 12, 2025 at 10:50 AM
The BBC published its first “AI assisted” bullet point summary.

Their press release said journalists will use an approved prompt for AI (they don’t say which) to generate the summary, then review and edit the output before publication.

They think young people like it.
www.bbc.com/news/article...
July 11, 2025 at 9:51 PM
Shocking story from @john-guinn.bsky.social on BBC today about a police Facebook post (which is still live) where the police say "Bring your extra strength alcohol along with you to St Peters Green... Illegally enjoy your beverage from one of the many benches"
www.bbc.com/news/article...
June 27, 2025 at 5:23 PM
"The AI native is an idiot… the AI native lacks critical skills… and that is terrifying… we have to stop the decline of decision making skills.”

Kristian Hammond, professor of computer science, Northwestern University at Chicago AI Week.
June 18, 2025 at 3:13 PM
What interests me is the past examples of bad use of AI in news that mainstream news reporters use to tell the story.

CNN’s list of offenders includes: CNET, Gannett, Apple Intelligence, LA Times and Bloomberg News.

edition.cnn.com/2025/05/28/m...
May 30, 2025 at 10:01 PM
“The freedom of our thoughts is one of our greatest assets, perhaps the greatest of all, and now it’s up to us to protect” Janosch Delcker ending his fascinating look into the future of mind reading AI at Republica in Berlin today.
May 27, 2025 at 10:17 AM
Conor Gogarty did some good research but he's working within a context where the readers are already tired of years of low quality content. The commenters on the FB post criticise Wales Online (which I guess the FB algo would see as engagement so push into feeds).
www.facebook.com/WalesOnline/...
May 19, 2025 at 12:35 PM
I just checked the domain - thenational.wales - and it’s giving me an error “404 page not found”.

So Conor may have done us all a favour and got rid of one source of slop polluting the information ecosystem.
May 12, 2025 at 5:12 AM
The @ebu.ch has proposed a news code of practice to AI developers and set out five core principles

www.ebu.ch/news/2025/05...
May 5, 2025 at 7:06 PM
@alexandraborchardt.bsky.social has made this really handy summary of research asking the audience what they think are acceptable and unacceptable ways journalists should use AI.
April 30, 2025 at 9:26 AM
Gard Steiro, Editor-in-chief of Norwegian news service VG shared at the Nordic AI summit that they have a tool which converts an article into a video.

www.youtube.com/watch?v=t80E...
April 29, 2025 at 6:20 PM
@chrismoranuk.bsky.social shared at #IJF25 that they didn’t implement their experiment in automating the writing of live page summaries partly because of how difficult for journalists to fact check AI generated content:

“It’s super hard to spot a single error in 400 words.”
April 12, 2025 at 12:42 PM