🧙🏼 ChatGPT's search improvements

SUBSCRIBE | ADVERTISE

Welcome to What’s brewing in AI—the newsletter that serves you only the most essential news, and practical ways you can use AI right now.

Howdy wizards,

A special welcome to the 220 new subscribers who joined last week.

In this edition:

• Search improvements in ChatGPT, and what it means for your business

• Visa and Mastercard are giving AIs their own credit cards soon

• How Claude and ChatGPT helped me optimise my ads campaign (continued)

• PSA: You now only have to bring three friends for free, lifetime access to contextwindows.ai. Details at the end.

I hope you’re seated comfortably with a cup of damn fine coffee.

Here’s what’s brewing in AI.

DARIO’S PICKS

1. Search improvements in ChatGPT

ChatGPT now does about 1 billion web searches a week. OpenAI rolled out several improvements to their web search this week, most notably product cards with visual details, pricing, reviews and—of course—a direct link to buy.

What’s improved (ranked by importance):

1. Actually usable shopping results. Chat now gives you better product results in cards. This matters more than it might seem for the user experience. OpenAI says products are chosen independently and are not ads—we’ll see how long that lasts.

2. Better citations. It can now give multiple citations for a given response; it also highlights the text if you hover over the source (👏 extremely useful).

3. Trending + autocomplete suggestions. You now see trending searches before typing and autocomplete suggestions (a lá Google search) as you start typing.

4. Search available in WhatsApp. You can now text 1-800-ChatGPT for up-to- date answers.

‎ Why it matters‎ ‎ This is a consequential piece of AI news that went under the radar to most this week. Take note—AI is rapidly becoming the middle-man between potential customers and your business.

SEO is moving to AIO (AI Optimisation) fast. With AI, it’s all about super-long-tail/detailed search terms, e.g. from “Best espresso machines 2025” to “best espresso machines under $200 that are great for making lattes and nice for small counter space”.

OpenAI is going head to head with Google with these improvements, as well as finally closing important UX gaps to Perplexity (well-known for its fast AI web search and visual results).

IN PARTNERSHIP WITH HUBSPOT

How can AI power your income?

Ready to transform artificial intelligence from a buzzword into your personal revenue generator

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

• A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential

• Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background

• Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

DARIO’S PICKS

2. Visa and Mastercard introducing credit cards for AIs

Visa is introducing Intelligent Commerce, a new platform for enabling AI assistants to handle shopping and payment on behalf of users. It’s a suite of APIs and a partner program that’ll allow developers to integrate Visa payments in their AI shopping experience.

Here’s what Intelligent Commerce means for each party:

• Consumers can soon pay directly from their AI assistant with one‑time digital tokens that you can set a spend limit on

• Developers of AI apps are getting a plug‑and‑play Visa API that handles payments and compliance

• Merchants are getting trusted, low‑friction checkouts backed by Visa’s fraud protection

Mastercard is also rolling out a similar platform called Agent Pay, which is starting out as more of a gated program, in collaboration with Microsoft and IBM before broadening access.

‎ Why it matters‎ ‎ AI’s influence in shopping decisions is growing exponentially—and initiatives like this from the major payment networks will help boost important parts of the experience including speed and safety, at scale.

UP CLOSE

In this mini-series I share different ways I’m using AI from week to week, as well as practical tips & tricks I discover and actually use.

Using ChatGPT and Claude as my Meta Ad managers, pt. 2

A few weeks ago I I had Claude help me set up a Meta ads campaign (here’s part 1)—my new strategy to drive even more subscribers (wizards) to this newsletter. I want to show you how I continue to optimise the campaign it helped me set up.

Rather than presenting myself as an expert (give me a few months first) or turning this into another Meta ads tutorial, my goal here is to show you how I think and experiment with AI as a partner in the work I do, so you can apply similar approaches to your unique challenges.

Setting the campaign up with Claude

Three weeks ago, as mentioned in part 1, Claude helped me with creative writing and setting up the structure for the campaign.

What I had Claude do:

• Copywriting for the ads (90% written, with some creative direction and tweaking by me)

• Deciding on campaign structure (budgets, ad sets, variations)

• Creating a naming convention for tracking

To compare, a few days later I decided to setup an additional identical campaign—same ad format but with the ads purely hand-written by me.

Performance so far

So far, both campaigns have settled around $2.7 average cost per new subscriber (CPA).

Optimisation with o3

To check in on my ad performance, I pasted screenshots from the Meta Ads dashboard showing performance across campaigns, ad sets, and variations.

I asked both Claude 3.7 Sonnet (with extended reasoning) and OpenAI’s o3 model inside ChatGPT to analyze the results and recommend optimization steps. I then compared their approaches to see which AI offered the more useful insights.

Claude did a good job identifying ad variants with high CPAs, and presented findings in straightforward text with bullet points (not optimal). It also suggested pausing underperforming ads, but here it had a blind spot; it recommended cutting ad variants with very little data, focusing solely on CPA and not considering the absolute number of coversions.

o3 did much better at breaking down my ad performance in an easy to digest way and giving me targeted tips. It managed to get all the data from all my messy screenshots right. This might be the new vision ability combined with the improved reasoning at work (I mean, o3 can solve Harvard Business School cases from PDFs, in one shot).

Contrary to Claude, it only made recommendations on optimisation for ad variants with a certain amount of data (ie it handled the complexity of the data better). It also made practical recommendations about how big changes I could make to existing ads while avoidign reentering Meta’s learning phase.

What really set it apart for me though, was that it presented the data much better (screenshot below); clever use of emojis, and the table formatting inside ChatGPT—good contrast, lots of data in the viewport at a single time but not too much—is also a big step above Claude.

Deciding on the next steps

Based on o3’s recommendation I ended up extracting the top 3 best performers and creating a new ad campaign based off just those. I also paused some underperforming ad sets.

I’ll leave the new campaign running for a while, and then I’ll try creating some fresh ad formats (with the same messaging) in a separate ad set later.

I’ve also discovered an amazing tool for spying on high-converting Facebook and TikTok ads for any company. You can see things like their longest running ads (ie their best performing ones). For the next campaign, I might borrow some creatives I like and have Claude it them adapt to my newsletter.

Takeaways

• Don’t forget who’s steering the ship. Y’know that zen quote about the mind being a wonderful servant and a terrible master? I think the same might be true for AI. You want a helper and a sparring partner, not something that dictates or makes decisions for you.

• Experiment with different AIs, but not too many. Have premium subscriptions for at least two major chatbots (ChatGPT, Claude, Gemini, Perplexity) so you can leverage reasoning models without the strict rate limits, and test multiple AIs on different parts of your workflow. I would advise against having every damn AI out there because it’s easy to end up focusing too much on the tool itself and too little on the most important thing: your problem—and why it needs to be solved.

• Get some skin in the game. This is more of a personal preference, but finding a use case for AI where the goal is to directly invest, save or earn money (even if just a tiny bit), makes it more fun to learn. You could try a similar approach to any type of paid ads or investments. Let AI analyse pattern and get a feel for when it makes the right calls and when it’s off.

  • Example: use it to try negotiate a better deal on some SaaS subscriptions you have, or have it analyse which subscriptions in your company are underutilized and can be consolidated/eliminated.

—

🙋🏻‍♂️ Are these practical AI stories useful to you?

I’m hungry! For your feedback.

CONTEXT WINDOWS

I want to lower the threshold for everyone to get a clear overview of how leading companies are implementing AI. My database is the best organised and most comprehensive view you’ll find. That’s why I’m slashing the requirement to get access for free to just three referrals (from six).

Stay ahead of your competition—drop your sharing link in your company Slack this week and you’ll have free access before you know it ↓

THAT’S ALL FOLKS!

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.