AI for marketing agencies: how to use it in your digital advertising campaigns

Artificial intelligence is already part of the daily work of anyone managing campaigns on Google Ads, Meta Ads and other platforms. It does not replace a specialist’s judgement, but used well it saves hours on repetitive tasks: building reports, reviewing thousands of search terms, writing scripts or producing ad variations. In this guide we cover four concrete uses of AI in campaign management, with examples you can apply today in your agency or your own business, and the precautions worth taking.

1. Campaign reports with Claude Design and your own design system

Claude Design is an Anthropic tool, launched in April 2026, that lets you create designs, presentations and visual documents by talking to Claude. It is available on paid plans (Pro, Max, Team and Enterprise), and your work can be exported to PDF, PowerPoint, Canva or HTML.

The most useful part for an agency is the design system: when you set it up, you upload your logo, fonts, and presentations or reports you already like, and Claude builds a colour palette, typography, components and page layouts. Once published for the team, every new report uses that same system automatically, so all your reports keep the same identity even if a different person builds them each month.

A simple workflow for the monthly report:

1. Set up the design system with your brand and real reports, not just the palette: Anthropic recommends including real examples because they produce better results.
2. Define a fixed structure: monthly summary, KPIs against the previous period, what worked, what did not and next steps.
3. Each month, export the platform data and ask Claude for the report with that structure.
4. Review figures and conclusions, adjust what you need and export it to PDF or PowerPoint for the client.

2. Data analysis with Claude and ChatGPT: the search terms example

Both Claude and ChatGPT can read CSV or Excel files and analyse thousands of rows in minutes. A very concrete case in Google Ads is the search terms report: the real searches that triggered your ads. Reviewing it by hand to find what to exclude takes hours; with AI, the job becomes one of review.

The key is giving it the right context. Download the report as a CSV, with cost, clicks and conversions, and add:

• What the business sells and, above all, what it does not.
• The areas it serves and the type of customer it targets.
• Examples of searches that work and of ones that never convert, such as “free”, “jobs” or “course”.
• What you want back: a list of terms to exclude, with the reason and the suggested match type.

An example request: “We are a real estate agency in Buenos Aires that only sells flats; we do not rent or do valuations. Attached are the search terms from the last 30 days. Tell me which ones you would add as negative keywords, with the reason, and do not include any term that has conversions.”

Turn the process into a skill

If you repeat this analysis every week, there is no need to explain everything again. In Claude you can create a skill: a folder with instructions, and scripts or supporting files if needed, that Claude loads when the task calls for it. Custom skills are written in Markdown, with no coding required, and are available on every plan with code execution enabled. In ChatGPT you can get something similar by saving the instructions in a project. That way each client’s context and your exclusion criteria are saved, and next time you only upload the file.

3. Google Ads and Microsoft Advertising scripts with ChatGPT and Claude Code

Google Ads and Microsoft Advertising let you automate tasks with scripts: JavaScript code written in an editor inside the platform itself. They can be scheduled to run once, daily, weekly or monthly, at the time you choose. Microsoft Advertising has no hourly option, but you can duplicate a script and schedule each copy at a different hour. In Google Ads, each run is limited to 30 minutes.

Writing a script used to require knowing how to code. Today you can ask ChatGPT or Claude Code to write it following each platform’s official documentation, explaining what you want to achieve. Some useful examples:

• Email alerts when a campaign spends faster than planned for the month.
• Detecting ads whose landing page returns an error and pausing them.
• Flagging keywords with spend and no conversions in the last 30 days.
• Sending daily metrics to a spreadsheet to feed a dashboard.

An important tip: start with scripts that only report before using ones that change the account, and always test them with the preview, which shows what the script would change without applying it. Microsoft especially recommends this because scripts have no undo button.

4. Ad design with ChatGPT Images 2.0 and GPT-6 Astra

OpenAI released GPT-6 Astra on 3 September 2026, available on ChatGPT Plus, Pro, Business and Enterprise. For design, its advantage is that it works together with ChatGPT Images 2.0: it can choose the image model, set the aspect ratio and use existing ads as a reference to write new variations. Images 2.0 also greatly improved how legible text is inside images, which is key for ads with a price or an offer.

In practice, from a single concept you can ask for versions for each placement:

• Square 1:1, for feeds.
• Vertical 4:5, for the Facebook and Instagram feed.
• Vertical 9:16, for Stories, Reels and TikTok.
• Landscape 1.91:1 or 16:9, for Display, Demand Gen and YouTube.

For better results, give it a full brief: brand, product, audience, offer and tone, and ask for variations of the same concept to test. Then check that the product looks accurate, that the text has no errors and that the piece meets each platform’s policies.

More ways to use AI in campaign management

• Ad copy: headlines and descriptions for responsive search ads, within the 30 and 90 character limits, grouped by theme or audience.
• Video scripts: hooks for the first seconds and a scene-by-scene script for TikTok and Reels.
• Product listings: clearer titles and descriptions for Merchant Center, Mercado Libre and Amazon catalogues.
• Keyword grouping: sorting a Keyword Planner export by intent to build ad groups.
• Account audits: spotting settings worth improving, such as broad match without smart bidding, missing negatives or poorly set locations.
• Competitor analysis: summarising the messages and offers your competitors use, from the Meta Ad Library and the Google Ads Transparency Center.
• Landing page versus ad: checking whether the landing page delivers what the ad promises and suggesting conversion improvements.
• Tracking: explaining GA4 events, writing regular expressions and code for Tag Manager or BigQuery queries.
• Adapting by country: tailoring the same campaign to Argentine Spanish, Spain’s Spanish or English, without literal translation.

Always review the output and protect the data

AI makes mistakes: it can misread a column, invent a figure, write a script with bugs or produce an image with an extra letter. Everything it produces is a first draft to review before it reaches the account or the client:

• Check a report’s figures against the platform.
• Test every script in preview before scheduling it.
• Review copy, product and policies in every image.
• Never exclude a search term that has conversions without analysing it.

Protect privacy too: do not upload clients’ or leads’ personal data, such as emails or phone numbers, anonymise exports where possible and check each tool’s data policy. For professional use, business plans are the better choice.

AI speeds up the work; deciding what to publish, what to exclude and where to invest is still a human job.

How we use AI at Semcial

At Semcial we have standardised processes that use artificial intelligence for reports, search term analysis, scripts and creative, always supervised by our account managers and designers. AI is there to speed up the work: it frees time from repetitive tasks so we can spend it on the strategy and optimisation of each account, and everything goes through human review before it reaches your campaigns.