Generate ads from Studio

Targeting writes one campaign brief for a web flow and opens the chat you already use, with that brief filled in. You review the instruction and send it. The chat drafts a paused ad. You turn it on.

This is for web flows. Message flows do not get an ad brief.

Connect the ad account first if you want conversions, spend, and the Live mark. That connection lives in Console, not in the chat. See Connect Meta, Google, LinkedIn, and ChatGPT Ads.

Open Targeting

1. Open the web flow in Studio.
2. Open Targeting in the section menu.
3. Pick Basis, Organic, Meta, Google, LinkedIn, or ChatGPT Ads.

The brief is created when you generate the flow with Targeting / Campaign brief checked. The label in the generator is “Campaign brief for Meta, Google, LinkedIn, and ChatGPT.” You can regenerate the flow later with that box still on.

What the menu marks mean

Mark	Means
Check on Meta, Google, LinkedIn, or ChatGPT Ads	That account’s connection status is active.
Live on an ad network	The account is connected, and a campaign attributed to this flow had impressions, spend, or clicks in the last 30 days.
Check on Organic	The share title and the share description are both filled in, and neither is leftover generation-prompt text. An image is not required.

If the account is not bound, the page says ads setup is not available. Connect it under Console → Settings → Integrations.

Basis and Organic

Basis is the company image library. Select the images the brief may use. Generate includes those keys in the instruction. If none are selected, the instruction says an image is still needed.

Organic is the share card for the flow link: title and description.

Generate

On Meta, Google, LinkedIn, and ChatGPT Ads the order is the same:

1. Setup — connection, account, pixel, conversion action. Fix anything listed there before you expect Live ads.
2. Bot and Generate — directly under Setup.
3. The settings the chat will use — primary text (introductory text on LinkedIn), headlines to test, descriptions to test, images to test, and the call to action.

Pick ChatGPT, Claude, Grok, or Gemini. Studio stores that choice on your user, so the next flow opens the same chat.

The line under the button says Generate opens that chat from the settings below. Review the instruction there and send it. The model is the one selected in that chat.

• Claude and ChatGPT leave the instruction for you to send.
• Grok may send it as soon as the chat opens.
• Gemini does not keep a prefilled question. Studio copies the instruction to your clipboard. Paste it into the chat.

Claude shows its own caution line above a prompt that arrived from a link. That line is normal. It is Claude checking the link, not a problem in the brief. ChatGPT opens the same instruction without that banner. Either way, you read it and send it.

The instruction tells the chat to draft one paused ad and not to publish it or turn it on. It also says to put the text variants in as separate versions to test, and to gather more assets from the website or ask you for them when it needs images with a call to action and text.

Google and ChatGPT have no ads connection inside the chat. The instruction tells it to use the browser, signed into Google Ads or ChatGPT ads, to create the paused ad, and still not to turn it on.

What is in the instruction

• Campaign name. The UTM key is minted once as smoc- plus the flow’s template key, and kept.
• Destination URL. It is the Flow URL with that network’s parameters already attached. The chat is told to use it exactly.
• Character limits for that network.
• Offer and audience from the brief.
• Headline variants and description variants.
• Primary text, or introductory text on LinkedIn. Google uses headlines and descriptions only.
• Image variants, when the brief has them.
• Selected Basis images.
• Regions to publish and Languages to publish, when the flow has them. An empty selection means every country, sales region, and language configured on the company. Cities from the generation settings are included on the regions line.

Why it is fast, and how the loop works

A blank chat is fast only after someone has typed the offer, the landing page, the limits, the countries, and the images. Studio has already done that. Generate opens ChatGPT, Claude, Grok, or Gemini with the brief filled in, so the first answer is the model's writing time, not a briefing.

Update is the feedback loop. Once ads on that network send people to this flow, Update opens the same chat with those ads and the recent results. The chat recommends the next variants from what delivered, not from a new blank prompt. You apply the ones you want. The destination stays the Flow URL, so the next round of clicks and leads still joins the same campaign.

The passes are:

1. Generate drafts paused ads from the brief.
2. You turn on the ads you accept.
3. Targeting shows Live when an attributed campaign delivers, and Console keeps the spend and the leads.
4. Update sends the live ads and the results back to the chat.
5. You test the next variants. Repeat.

Update

Update is shown only when that network already has at least one ad sending people to this flow.

It opens the same chat. The instruction lists each connected ad and asks for one recommendation per ad: what to change, new variants to test, and why the current results support the change. It still says not to publish or turn the ads on. Google and ChatGPT are told to apply each change in the browser and leave the ads off.

There is no Update button when no ad is connected.

After the chat drafts the ad

1. Read the draft in the chat.
2. Leave it paused until you are ready.
3. Confirm the destination is the Flow URL and the URL parameters match the recipe on the integration page.
4. Turn the campaign on in Ads Manager, Google Ads, Campaign Manager, or ChatGPT ads when you want delivery.
5. Come back to Targeting. After the next sync, a delivering campaign can show Live.

Related

• Connect Meta, Google, LinkedIn, and ChatGPT Ads
• Track conversions with Head Tags (fallback)
• Embedding web flows and passing UTMs
• Generating a flow with AI in Studio