Vol. I · No. 044Saturday, August 1, 2026Free edition
The Memo

Where marketing meets AI, decoded every morning

Ad PlatformsBriefing

ChatGPT prices ads, YouTube draws a line: July 20 brief

ChatGPT publishes ad bids, YouTube tightens its AI video line, and Kimi brings an AI agent to the desktop in the July 20 business brief.

The MemoJuly 20, 20264 min read
ChatGPT prices ads, YouTube draws a line: July 20 brief

Some links below are partner links. If you buy through one, The Memo earns a commission at no extra cost to you. How we make money.

In this briefing

ChatGPT has put a price marker on its ads, giving small advertisers their first useful budgeting signal. The change matters to businesses testing new places to reach buyers. YouTube is also spelling out which AI-made videos can lose ad revenue, while Kimi is asking owners to trust an AI agent with their desktop.

ChatGPT puts a number on ad bids

OpenAI updated its ChatGPT Ads guidance on July 19 with a concrete starting range: $3 to $5 for each click. Advertisers can pay per click or per thousand ad views. Ads appear below relevant conversations, and OpenAI says its system considers the conversation, the landing page and the ad itself when deciding what to show. It also accepts “context hints,” meaning descriptions of the topics where an offer belongs, instead of relying only on search words.

The official buying guide says reporting covers views, clicks, spend and conversions, which are completed actions such as a purchase or inquiry. That makes ChatGPT Ads measurable enough for a controlled test. Yet OpenAI has not published broad performance benchmarks. A $5 click is only cheap when it produces profitable customers.

For local services, travel and household products, this is a real new test channel. OpenAI's July advertising policy says those areas are supported during the initial period, while some financial, health and legal advertisers face added review. Businesses in restricted categories should wait for clear eligibility instead of building a campaign that cannot run.

Your move

Set a small fixed test budget and send ChatGPT ad traffic to one page with one measurable action. Compare the cost of each qualified inquiry with your existing search ads after 30 clicks. Stop if the economics miss your target.

This fits a wider shift we covered in how AI assistants recommend businesses. Buyers are making choices inside answers. Paid placement now sits beside that discovery, but it does not replace accurate product details, clear reviews or a page that earns trust.

YouTube makes low-value AI video a revenue risk

YouTube clarified on July 20 that channels built from interchangeable, manipulative or shock-led videos can lose access to ad revenue. Its monetization policy now gives a direct allowance for AI-assisted work when the finished video shows a creator's own idea, research or point of view. Generic AI templates produced at scale sit on the wrong side of that line.

This affects any business using cheap video volume to grow a channel. Losing monetization is one cost. A library of near-identical clips can also weaken customer trust before a viewer ever reaches the offer. AI remains useful for scripting, editing and visuals. The deciding factor is whether a person added judgment that viewers can recognize.

Audit the newest and most-watched uploads first, because YouTube says reviewers may focus there. Remove repeated story patterns, add firsthand explanation and make each video answer a distinct customer question. Our small-brand AI video guide can help separate production tools from the creative work a person still needs to own. The safer standard is simple: automation can build the clip, but your business must supply the reason to watch.

Kimi gives an AI agent access to the desktop

Kimi introduced Kimi Work on July 16 for Mac and Windows, moving its assistant from a web chat into local folders and a signed-in browser. The company says the desktop app can read and write files, click through web pages, run scheduled jobs and coordinate as many as 300 smaller AI workers on one assignment. An AI agent is software that carries out several steps toward a goal instead of returning one answer.

According to Kimi's product announcement, owners can choose full access or require approval for sensitive actions. That permission choice is the business story. A tool that can complete forms and change files can save hours, but a wrong instruction can travel much farther than a bad chatbot reply.

Small teams should ignore the 300-worker headline for now. Start with a reversible task, such as turning copies of weekly reports into a summary. Keep approval turned on, use a separate folder and compare every output with the source files. That same review habit applies to the AI writing tools already used by small teams. More autonomy raises the value of a clear check.

Worth watching

  • OpenAI says ChatGPT Ads uses an auction, so the suggested $3 to $5 click bid may move as more advertisers arrive.
  • YouTube's wording gives creators examples, but channel-level enforcement will show how aggressively it treats repeated AI formats.
  • Kimi says local files stay on the computer, while browser tasks can use signed-in sessions. Permission controls deserve close scrutiny before wider use.

About The Memo

The daily brief on AI and marketing. What changed in AI tools, search, ads and growth, why it matters, and the move to make this week. How we work

The daily Memo

One short email each morning when something in AI or marketing actually changes, with the move to make.

Free. Every morning. Unsubscribe anytime.

Keep reading

Related briefings