We recently passed the 100-day “anniversary” of Ask INMA, the first press association AI answer engine and the only one trained on our one-of-a-kind content universe. 

I want to update you on the impact of and insights from Ask INMA. It is a game-changer for leveraging the value of INMA membership — beyond any tool ever introduced in our 95-year history. Ask INMA users are quickly discovering the Ph.D. level discovery and insights possibilities — good for their companies, professionally good for them.

Since the launch on June 15, hundreds of members have made thousands of inquiries. Many are making Ask INMA central to their professional lives.

What members are asking

Across those queries, three dominant subjects emerge: 

• Subscription and retention strategy: 24% of questions

• AI and technology: 22% of questions

• Audience growth and engagement: 17% of questions

Burrowing into the nature and tenor of the questions, four insights emerge:

• There is a shift from “what” to “how”: Members are no longer asking for definitions. They are asking for implementation guides, benchmarks, and ROI models. This signals news organisations are actively investing in and executing strategies around AI, dynamic paywalls, and audience engagement. They require guidance on optimisation and scaling.

• AI urgency is focused on efficiency and ROI: The most frequent queries relate to automating workflows (transcription, summarisation, tagging), and improving newsroom productivity. There is growing pressure on leaders to prove the ROI of AI investments. In some ways, the AI focus on efficiency gleaned from Ask INMA questions goes against the human narrative that we’ve heard at INMA events around a productivity focus with AI. 

• The battle for direct audience relationships is on: Recurring questions about the "collapse of social media referrals" and the need to "boost direct traffic" point to a significant strategic pivot across the industry. Publishers are actively seeking to reduce their dependence on third-party platforms by investing in owned channels like newsletters, news apps, and community hubs. 

• Subscriptions and AI are becoming intertwined: While categorised separately, the topics are beginning to merge. Members are asking about “AI-driven paywall models,” “using AI to increase subscriptions,” and “AI for programmatic ad optimisation.” This indicates that the next frontier is not just using AI for content but integrating it deeply into the core revenue-generating parts of the business. 

INMA members are moving past theoretical questions on these subjects and are demanding practical, implementation-focused guidance on both fronts.

Ask INMA, a backstory 

It was not our original plan to publicly launch in June. Since we began conceptualising the AI in September 2024, our plan was to have Ask INMA as an internal tool for our staff and initiative leads to help members. Then we would launch publicly in four to six months. 

Through the process of implementation, our developers (Techlabee.ai) put me through questions I didn’t initially understand. What is the INMA voice? What is INMA style? Narratives vs. bullets? How can we report out answers that are unique to INMA? I never realised so much creative thought went into how and what we addressed in a query. 

The answers to these questions spoke volumes about what differentiates INMA from all other organisations: We are hopeful. We are practical. We are to the point. We are truly global. We are a community with whom we have direct relationships. We speak in case studies. We cite chapter and verse. We give the whole answer and don’t dump you to read more (unless you want that). We provide an alternative point of view. 

The results were amazing. And surprising.

The “problem” with our timetable quickly became apparent: Ask INMA was so damned good. Our initiative leads, notably Readers First Initiative Lead Greg Piechota, couldn’t break it. As we poked and prodded the AI in the second quarter, we all came to the same conclusion: Why wait? Why not bring INMA members along for the ride, launch publicly in beta mode, and iterate together?

Like many INMA members, creating Ask INMA was an exercise of learning about ourselves. We generally knew most INMA content was not searchable, but we had never run the numbers:

• Best Practice archive: 8,185 articles 

• HTML crawler: 8,092 files

• Blogs: 5,674 articles

• Presentations: 1,825 PPTs/PDFs 

• Video: 365 recordings 

• Audio: 199 recordings 

• Reports: 75 PDFs 

In INMA Search, you could peruse what we databased and HTML, but you couldn’t penetrate PDFs, PPTs, or audio and video recordings. AI changed that and opened up possibilities we could only dream of. We learned that 91% of INMA content could not be discovered by public search engines and answer engines — and we remain OK with that. 

INMA has maintained its Search capability. While Search has declined slightly, it remains alive and well because the Ask INMA answer engine is only available to members.

Ask INMA, suddenly, became the portal to nearly all INMA content.

By the numbers: sources, upgrades, usage 

And INMA members have come along for the ride: 

• Since launch, 878 INMA members have made 4,529 queries.

• 330 of those members are returning users — 10% of whom are superusers. 

• The average session length is six minutes and seven seconds. 

• The average answer time is 102 seconds. Ask INMA’s default is “deep research mode,” thus answers take longer than ChatGPT and equivalent. 

• On June 15, we had indexed 8,638 sources. Today, that has expanded to 33,171 sources.

• We have made (and chronicled) 92 updates, improvements, and fixes to Ask INMA in its first 100 days.

What’s next 

This blog is designed to update INMA members on “the ride” with Ask INMA. 

We remain in beta mode. We’re getting member feedback. We’re constantly learning. We continue to iterate and improve. 

I love learning from how INMA members are engaging with the answer engine. For example, being patient and asking the same question multiple times seems to get better answers. Providing persona-like context in the prompt. Not necessarily being polite: If you don’t get the answer you want, ask again … be pushy. Having said that, I’m amazed at how some members are unflinchingly polite to Ask INMA (“thank you,” “please,” “good answer,” “appreciate you”).   

AIs are imperfect, yet the good dramatically outweighs the bad — the worst for INMA being less about hallucinations than omissions. 

I hope that you will use Ask INMA — and encourage your teams to do so, too. As we move forward in beta mode, I also hope you’ll come along for the ride on this fascinating INMA journey.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .