“Frictionless” Content Creation: The Pros and Cons of Using AI for Content

In this exploration of what AI can and cannot do for content work, Troy Thibodeaux, AP’s director of AI research and adoption, will share use cases from creative fields that show how and when to use AI for content, as well as exactly what not to do. He’ll cover practical AI use cases for content teams and the ways in which the rules for AI use may vary according to the medium (text, photography, video, interactive graphics etc.) or a person’s line of work. “AI tools are focused on reducing friction in our workflow,” says Troy Thibodeaux. “If we let AI command the process, it will become frictionless, but also thoughtless. And we pay the price for that.” 

The session will include some tools demos, as well as instructions for prompting for better, safer results. There will be ample time for audience Q&A.

Why Standards Matter for AI Use in Content Work

Join Amanda Barrett, AP’s vice president of news, standards and inclusion, for an inside look at how the world’s largest news organization approaches standards and guidance for AI use. She’ll cover the red, yellow, green zones of when to avoid AI, when to use it with caution and when to go for it. She’ll also share how the Associated Press uses AI in its operations, how it oversees quality control over AI outputs and how it ensures a healthy layer of editorial judgment. This interactive session will also cover transparency, disclosure and standards – and how disclosure can be a double-edged sword. The session will close with a look at how guidance on AI use can vary with each content format: text, photography, video and illustration. There will be time for audience Q&A. You’ll leave with ideas on how to establish standards for your personal content work or for your organization. 

Introduction to Authenticating Online Content

The new verification problem 

In 2026, verifying user generated content (UGC) has shifted from identifying crudely generated AI video clips and images to assessing highly plausible synthetic and hybrid media. Both audiences and people working with content now encounter a mixture of generative AI images and videos, cloned voices, altered documents and genuine footage presented with false captions. “Real or fake?” is often the wrong opening question. News consumers should consider who created the content, when and where it originated, whether it was altered, and whether it is consistent with the accompanying claim. 

Common types of synthetic and manipulated content 

Experts from the Associated Press user generated content verification team will present examples of manipulated content - from fully synthetic media to subtler alterations. We will also examine “cheapfakes,” because misleading context, selective editing and false captions can be just as deceptive and consequential as sophisticated deepfakes.

Red flags and first-pass checks 

You will learn how to spot red flags as reasons to investigate, not proof of fabrication. We will demonstrate visual indicators such as unstable text and inconsistent reflections or shadows, as well as audio indicators such as distortion or unnatural pacing. We will discuss reverse image search, the cornerstone of the verification process, and the value of identifying original content sources.

Verification, Authentication Tools and Editorial Conclusions

Verification workflow 

Experts from the Associated Press user generated content verification team will teach a repeatable, evidence-led workflow for verification. We will show you how to investigate the source and chain of custody, trace the earliest available upload, reverse-search images and video keyframes, and verify time and location through landmarks, shadows, weather, language and independent reporting. 

Tools for authentication 

You will learn how LLMs like ChatGPT or Google Gemini can be used for verification workflows. We will introduce C2PA Content Credentials and SynthID watermarks. We will explore how commercial AI detectors and ChatGPT or Gemini workflows should be treated as fallible investigative leads, not editorial verdicts. 

Communicating uncertainty 

You will learn what to do when absolute verification isn’t possible. We will show you how to report what the evidence supports without overstating certainty, and when to describe content as “verified”, “confirmed synthetic” or “inconclusive”. You will learn to avoid declaring content “fake” solely because a detector produced a high score or because metadata is absent.