ChatGPT Ads wants your CRM, Google ships Argon, Claude goes to the bank

ChatGPT Ads now takes your first-party data. According to Search Engine Journal, citing The Information, LiveRamp's RampID lets advertisers reach or suppress audiences built from CRM systems, loyalty programs, websites and apps, starting in 11 markets. LiveRamp has measured ChatGPT ad conversions since June; now it handles targeting too. OpenAI's Ads Manager also accepts customer list uploads directly and lets you bid from 0.1x to 10x on an audience at the ad group level.

Then Digiday asked agencies how the ads are doing, seven months after OpenAI started selling them. Accuracast, running a UK lead generation campaign, saw about 100 clicks on the platform against 20 in its own tracking, and form fills arriving while ChatGPT still showed zero conversions two weeks later. Web Guide Partner sees conversions show up 24 to 36 hours late, against a few hours on Google. Not everyone sees it: AdRoll, Jellyfish and an independent US agency told Digiday they found no click gaps. OpenAI declined to comment. One executive described clients who spend $10 million a month on Google and under $100,000 on ChatGPT.

On September 30, Koray Kavukcuoglu of Google DeepMind introduced Gemini 4 Argon, Google's new frontier model. It goes first to trusted cyber defenders through Google's Fairwind Program. Paid API customers and Google AI Ultra subscribers come next, on a schedule Google describes as "as soon as possible." TechCrunch notes it is only rolling out to a select group of Google's cyber partners for now.

Most of the numbers are Google's own: 77.9% on DeepSWE v1.1, first place on Zapier's AutomationBench at 51.3%, and an output limit raised from 64K tokens to 1 million. The introductory API price is $2 per million input tokens and $10 per million output tokens; the regular price Google lists is $4 and $20. The independent check comes from Artificial Analysis, which scores Gemini 4 Argon (high) at 53 on its Intelligence Index, level with GPT-6 Astra (max) and behind Claude Opus 5.5 (max with fallback) at 58. Our favorite line in the post: a team of Argon agents found memory savings across Google's data centers that will free more than 300 TiB once rolled out. The rest of us are still deleting vacation photos to free up 4 GB.

Marketing insight: Skip the leaderboard fight and watch the automation score. Zapier's AutomationBench tests business workflows, the kind your team runs in spreadsheets and inboxes. Before Argon reaches paid accounts, write down three repetitive workflows, such as lead routing, the weekly report and product feed updates, with the minutes each takes today. When it ships, you can test it against your current tool in an afternoon. And budget at the regular $4 and $20, so the second invoice holds no surprises.

Anthropic said on October 1 that Barclays is expanding its use of Claude. In the bank's Global Markets business, Claude sorts about 120,000 client emails a day. An internal assistant built on Claude has more than 16,000 staff users and has handled over a million searches. Barclays expects half of its developers on Claude Code by the end of 2026 and most of its engineers by 2027. The figures come from the two companies, not an outside audit.

A day earlier, Claude for Government became generally available to US federal and state agencies, running in a FedRAMP High environment. The pricing is the clever part: no seat fees, usage bought in fixed increments, a hard not-to-exceed cap and alerts before the balance runs low. Admins set spend and model limits for each user group.

Every story here comes down to one question. Can anyone check the work? Advertisers are holding ChatGPT at test budgets until the answer is yes. Anthropic answered it for government buyers with a cap and an alert. In our view, the companies that make their numbers easy to verify will get the bigger budgets. Models or mattresses, same rule.

If you want help building ChatGPT Ads tracking you can trust, or an inbox that sorts itself, that is the exact conversation our founders like having. Talk to them below. See you in the next edition.