Aisvarya Chandrasekar
LightNews LightNews
banner
aisvarya17.bsky.social
Aisvarya Chandrasekar
@aisvarya17.bsky.social
140 followers 330 following 16 posts
I work at @towcenter.bsky.social as a computational research fellow, studying the impact of AI tools in journalism. I am a grantee in Brown Institute’s 2024-25 Magic Grant cohort, building PollFinder.ai.
Posts Replies Media Videos
Pinned
aisvarya17.bsky.social
Aisvarya Chandrasekar @aisvarya17.bsky.social · Mar 6
@klaudia.bsky.social and I spent a few months studying how generative search tools cite news content. We'd love to hear your thoughts on our findings!

@towcenter.bsky.social @columjournreview.bsky.social
aisvarya17.bsky.social
@technologyreview.com blocks all of OpenAI's and Perplexity's disclosed web crawlers in their robots.txt. Despite this, both OpenAI’s Atlas and Perplexity’s Comet retrieved the full text of a subscriber-exclusive article when asked to. How?

Read @klaudia.bsky.social and my recent CJR article:
October 30, 2025 at 3:07 PM
Reposted by Aisvarya Chandrasekar
emilybell.bsky.social
In a shock development - AI companies are failing to benchmark what matters in newsrooms. More excellent work from our Tow Center research team @klaudia.bsky.social and @aisvarya17.bsky.social
columjournreview.bsky.social Columbia Journalism Review @columjournreview.bsky.social · Sep 18
The performance tests used by AI companies don’t measure what matters in the newsroom. By @klaudia.bsky.social and @aisvarya17.bsky.social for @towcenter.bsky.social. www.cjr.org/analysis/jou...
September 18, 2025 at 2:49 PM
Reposted by Aisvarya Chandrasekar
klaudia.bsky.social
In our latest piece for CJR, @aisvarya17.bsky.social and I argue that third-party evaluation of AI models is not just a technical matter; it’s a matter of accountability. www.cjr.org/analysis/jou...
Journalists need their own benchmark tests for AI tools.
The performance tests used by AI companies don’t measure what matters in the newsroom.
www.cjr.org
September 18, 2025 at 1:58 PM
Reposted by Aisvarya Chandrasekar
columjournreview.bsky.social
The performance tests used by AI companies don’t measure what matters in the newsroom. By @klaudia.bsky.social and @aisvarya17.bsky.social for @towcenter.bsky.social. www.cjr.org/analysis/jou...
September 18, 2025 at 12:08 PM
Reposted by Aisvarya Chandrasekar
fstegers.bsky.social
The Tow Center has looked into how good LLMs are at verifying and geolocating an image. Gentle reminder: "on multiple occasions models blatantly lied about the steps they took to arrive at an answer to our query" www.cjr.org/tow_center/w... #ai #osint
August 29, 2025 at 3:01 PM
Reposted by Aisvarya Chandrasekar
towcenter.bsky.social
We asked seven different AI models to identify the location, date and source of photos taken by photojournalists, they collectively only answered 14 out of 280 queries completely correctly.

@aisvarya17.bsky.social @klaudia.bsky.social @columjournreview.bsky.social

www.cjr.org/tow_center/w...
Why AI Models Are Bad at Verifying Photos
“You don't know when it's just making stuff up.”
www.cjr.org
August 28, 2025 at 5:12 PM
aisvarya17.bsky.social
New Research: Can AI models be used to fact-check images?

www.cjr.org/tow_center/w...

I cannot summarize this piece better than @klaudia.bsky.social did in this thread. The methodology and data from the experiment are linked in the article. Reach out with any questions, comments, or feedback!
August 28, 2025 at 12:45 PM
Reposted by Aisvarya Chandrasekar
klaudia.bsky.social
NEW in @columjournreview.bsky.social: Why AI models are bad at fact-checking photos www.cjr.org/tow_center/w...
August 26, 2025 at 8:12 PM
Reposted by Aisvarya Chandrasekar
dmil.bsky.social
Great work by @aisvarya17.bsky.social and @klaudia.bsky.social. Read more below!

www.cjr.org/tow_center/w...
Why AI models are bad at verifying photos
Many of the latest large language models, including OpenAI’s recently released GPT-5, tout an impressive capability: reasoning about images. We have seen viral trends of AI models geolocating obscure ...
www.cjr.org
August 26, 2025 at 10:25 PM
Reposted by Aisvarya Chandrasekar
foeke.bsky.social
More testing of LLMs for journalism, this time on identifying image location, date, and source.
By @aisvarya17.bsky.social and @klaudia.bsky.social with some of my own commentary.
www.cjr.org/tow_center/w... @columjournreview.bsky.social
August 28, 2025 at 9:17 AM
Reposted by Aisvarya Chandrasekar
columjournreview.bsky.social
The Future of News and Search:
The Tow Center for Digital Journalism interviewed news and tech industry representatives about AI’s impact on platforms and publishers. They expressed some hope and a lot of trepidation. By Klaudia Jaźwińska. www.cjr.org/the_media_to...
A New Report Takes On the Future of News and Search
The Tow Center for Digital Journalism interviewed news and tech industry representatives about AI’s impact on platforms and publishers. They expressed some hope and a lot of trepidation.
www.cjr.org
May 15, 2025 at 2:39 PM
Reposted by Aisvarya Chandrasekar
sparcopen.bsky.social
SPARC has invited @klaudia.bsky.social & @aisvarya17.bsky.social (Tow Center for Digital Journalism at Columbia U) @davehansen.bsky.social (Authors Alliance) & @petersuber.bsky.social (Harvard U) to discuss AI issues on May 29 at 1 pm EDT. SPARC members can register at: sparcopen.org/event/sparc-...
SPARC Webcast: OA, Attribution, and Addressing Faculty AI Concerns - SPARC
As AI adoption accelerates, authors have growing concerns over maintaining control of their own work that may reduce their willingness to share their research openly. Publisher licensing deals with AI...
sparcopen.org
May 20, 2025 at 2:05 PM
Reposted by Aisvarya Chandrasekar
klaudia.bsky.social
I wrote about the report - and publishers' anxiety around the decline of traditional search - in today's @columjournreview.bsky.social newsletter www.cjr.org/the_media_to...
Journalism Zero: How Platforms and Publishers are Navigating AI
<p>Download the PDF here. This research is generously funded by the Tow Foundation and the John D. and Catherine T. MacArthur Foundation. Introduction Early in our first interview, a veteran news exec...
www.cjr.org
May 15, 2025 at 1:19 PM
Reposted by Aisvarya Chandrasekar
klaudia.bsky.social
A new report I co-authored for the @towcenter.bsky.social is out today! We interviewed representatives from the news and technology industries about AI’s impact on the relationship between platforms and publishers. They expressed some hope and a lot of trepidation. www.cjr.org/the_media_to...
A New Report Takes On the Future of News and Search
The Tow Center for Digital Journalism interviewed news and tech industry representatives about AI’s impact on platforms and publishers. They expressed some hope and a lot of trepidation.
www.cjr.org
May 15, 2025 at 1:19 PM
Reposted by Aisvarya Chandrasekar
ndiakopoulos.bsky.social
Google's "AI Mode" in action. To the extent anyone uses it, this will be a traffic killer for news: two clicks to actually visit a site and you have to click the link icon for the sidebar to update with the sources so it also limits source credibility signals.
April 2, 2025 at 8:13 PM
Reposted by Aisvarya Chandrasekar
reutersinstitute.bsky.social
2️⃣ At 10:15 GMT we'll examine the power dynamics emerging between news organisations and AI developers, focusing on content licensing and data valuation

Featuring @fedecherubini.bsky.social @felixsimon.bsky.social @agstrait.bsky.social @mattrogerson.bsky.social @klaudia.bsky.social
buff.ly/dEmIIrg
March 17, 2025 at 10:10 AM
Reposted by Aisvarya Chandrasekar
columjournreview.bsky.social
Google and OpenAI want Trump to open up the rules. News publishers have some thoughts, by Klaudia Jaźwińska. www.cjr.org/the_media_to...
The Battle over AI and Copyright Enters a New Phase
Google and OpenAI want Trump to open up the rules. News publishers have some thoughts.
www.cjr.org
March 20, 2025 at 3:30 PM
Reposted by Aisvarya Chandrasekar
parismarx.com
Columbia Journalism Review tested eight generative AI search tools and found their answers were wrong 60% of the time, and the paid ones actually fared worse than the free ones.

Meanwhile, millions of people trust the way they present total bullshit with confident language.
AI search engines cite incorrect sources at an alarming 60% rate, study says
CJR study shows AI search services misinform users and ignore publisher exclusion requests.
arstechnica.com
March 18, 2025 at 3:24 PM
Reposted by Aisvarya Chandrasekar
emmbadger.bsky.social
Imagine if all of government reported data like DOGE does -- deleting numbers, altering them without notice, pointing numbers to the wrong things.

w/ another awesome @ethansinger.bsky.social graphic:

www.nytimes.com/2025/03/13/u...
See How Elon Musk’s Team Inflated, Deleted and Rewrote Its Savings Claims (Gift Article)
The cost-cutting group removed hundreds of contracts from its “wall of receipts,” added back many of them, and inflated savings values.
www.nytimes.com
March 13, 2025 at 8:51 PM
Reposted by Aisvarya Chandrasekar
emilybell.bsky.social
On a darker note - the hyper partisan dark money funded ‘news’ websites we have studied for 5 years or more @towcenter.bsky.social are taking an ugly AI turn. This was always going to happen - will not be the first or last incidence . www.nytimes.com/2025/03/12/u...
Yale Scholar Banned After A.I. News Site Accuses Her of Terrorist Link
The deputy director of a liberal project at Yale Law School was put on leave over allegations that she is linked to Samidoun, a group the U.S. government has said funds terrorists.
www.nytimes.com
March 13, 2025 at 4:36 PM
aisvarya17.bsky.social
It's heartening to see that our research is being shared widely and there is interest in holding AI tools accountable. But there have been some representations of our research that we would like to address.

www.cjr.org/tow_center/w...

@klaudia.bsky.social
AI Search Has A Citation Problem
We Compared Eight AI Search Engines. They’re All Bad at Citing News.
www.cjr.org
March 13, 2025 at 6:40 PM
Reposted by Aisvarya Chandrasekar
klaudia.bsky.social
This post is misleading. We were testing specifically to see whether the chatbots accurately identified the sources of excerpts from news articles. We did not intend to extrapolate these findings to the overall accuracy of the chatbots.
March 13, 2025 at 4:40 PM
Reposted by Aisvarya Chandrasekar
timnitgebru.bsky.social
Lets replace search with "AI" then! Totally logical if you ask me. Even more worth it when you know they're exponentially overtaking the airline industry in their carbon footprint.

Study: www.cjr.org/tow_center/w...
Tweet by Gisele Navarro @ichbinGisele "“Collectively, the AI search engines provided incorrect answers to more than 60% of queries.” ~ @CJR Screenoshot of an image showing "Generative search tools were often confidently wrong in our study" showing baplots of different offerings like ChatGPT, Perplexity and Deepseek, and showing the percentage of wrong answers in red and right ones in green.
March 12, 2025 at 9:34 PM
aisvarya17.bsky.social
Thank you for featuring our research!
niemanlab.org Nieman Lab @niemanlab.org · Mar 10
AI search engines fail to produce accurate citations in over 60% of tests, according to new Tow Center study
buff.ly
March 11, 2025 at 2:22 PM
Reposted by Aisvarya Chandrasekar
ryanbeckwith.bsky.social
well yeah, you gotta pay more for the confidently incorrect. regular incorrect is for freemium users
karlbode.com Karl Bode @karlbode.com · Mar 11
Tow Center for Digital Journalism study found that "AI" chatbots provided incorrect answers to more than 60 percent of queries, with Musk's Grok 3 responding to 94 percent of the queries incorrectly.

"Premium chatbots provided more confidently incorrect answers than their free counterparts."
AI Search Has A Citation Problem
We Compared Eight AI Search Engines. They’re All Bad at Citing News.
www.cjr.org
March 11, 2025 at 1:35 PM