PlaybookCampaign SetupReferenceCheatsheet

Gemini + Google Ads Workflow Cheatsheet: RSA, Demand Gen, QA

A Gemini-to-Google-Ads operating sheet: input pack, RSA character maths, Demand Gen asset copy, pin rules, policy QA and the upload order so generated text does not break the auction.

Updated 2026-08-2411 min read

Gemini can draft Google Ads assets faster than a media buyer can type. It cannot count characters, pin headlines, or know your conversion action. This sheet is the workflow: input pack, RSA maths, Demand Gen copy, pin rules and a QA gate before anything hits Google Ads.

Google Ads is an asset assembler, not a place to paste a paragraph

Responsive search ads mix headlines and descriptions you did not pair. Demand Gen mixes images, videos, logos and short text across Shorts, in-feed, Discover and Gmail. If Gemini returns a 90-character 'headline' or a poetic brand story, the UI will truncate it or the auction will assemble nonsense. Your job is to force Gemini into the real field lengths and to mark which lines may never sit next to each other. The model is a clerk for volume. The pin table and the character table are the director. Skip them and you will debug 'Gemini is bad at ads' when the actual failure is a 41-character Demand Gen headline or three RSA pins that freeze the ad into one weak combination.

Run one campaign type per prompt, then one QA pass

A prompt that asks for 'all Google Ads copy' will blend RSA 30-character headlines with Demand Gen 40-character headlines and Performance Max long text into one pile you cannot upload. Sequence: pack the facts, generate RSA for Search, generate Demand Gen text for the asset group, generate sitelink/callout extras only if those extensions are in scope, then run the policy and character QA. Do not let Gemini invent keywords you will not negate, prices that are not on the landing page, or callouts for a guarantee the PDP does not honour. Feed landing-page truth, not the brand deck.

Creative files are a different pipeline

Gemini text does not replace 9:16 video. Demand Gen without vertical video is a Display-shaped campaign wearing a new name. Generate the copy pack here, generate the video set elsewhere, then assemble in the asset group. Klip Kanvas covers the 9:16 / 1:1 / 16:9 file tree from a product URL; Gemini covers the headlines, descriptions and RSA combinations that sit next to those files. Upload order in the locked sheet is text-complete → files-complete → audience signals → bid. Reversing that order is how you launch a campaign that learns on a logo and one landscape clip.

How to use the tables below

Copy the input pack into Gemini once per SKU, with the campaign-type flag set so RSA and Demand Gen never share a prompt. Use the RSA table for Search character maths and mix, the Demand Gen table for asset-group text, the pin and never-pair rules before you lock any headline, then the QA and upload-order table as a launch gate. Each value column is a length, a count, a pin rule or a fail condition — not a pep talk about AI. If a line exceeds the cap, shorten it in Gemini and re-count; do not paste over-cap text into Google Ads and hope truncation is stylish. Video files still come from a separate pipeline; this sheet only clears the copy that sits next to them.

5 lookup tables inside: Gemini input pack per SKU, RSA field maths and headline mix, Demand Gen text character specs, pin and never-pair combination rules, then policy QA plus the six-step upload order.

Locked

Unlock the full cheatsheet

Enter your email and we'll instantly unlock the complete cheatsheet — plus send you fresh ad playbooks every month.

No spam. Unsubscribe anytime.

Ready to put this into practice?

Create your first AI UGC video ad in minutes — no filming, no actors, no editing.

Try Klip Kanvas free

More in this section

Ready to make ads like these?

Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.