How to use Claude for marketing

Claude for marketing is using Anthropic’s Claude to finish one named marketing job, usually research plus a campaign brief you can judge, then keeping a human as the judge on send, publish and spend. You start in Chat. You do not start by building a stack.

I am Adam Stewart. I run Digital Bond, and I write for founder-led Australian businesses that want AI to assemble the work, not to run the company.

This is how to use Claude for marketing this week. It is not a model review. It is not a bake-off. It is one loop you can run in Claude Chat.

If you want the three-jobs map (research, drafts, one follow-up loop), that sits on how to use AI for marketing.

How do I use Claude for marketing without building a stack first?

You open Claude at claude.ai. You stay in Chat.

You do not start with Claude Code. You do not start with connectors, plugins, or a new CRM. You do not build a marketing tool from a ChatGPT chat. That is a different job, for when the process is already named.

The loop is simple.

You dump what you already have: who you sell to, what you sell, one offer, the last campaign, a few real customer notes. Claude turns that into a campaign brief you can argue with. You remain the judge. Nothing gets sent.

If you cannot say the sentence for the loop, stop and read you don’t have an AI problem. Claude will not invent a system you have not named.

A working pass looks like this.

1. Paste the facts you trust.
2. Ask for a one-page brief: audience, offer, one channel, one next action, what must stay human.
3. Cross out anything you would not say out loud to a customer.
4. Hand the brief to a person, or to the tool that already sends email.

That is the job. A finished brief. Not a new stack.

Try this:

I run an Australian [service / ecommerce] business. We sell [offer] to [who it is for] in [city or situation].

Here is what I trust:
- Our current offer, in our words: [paste]
- Who bought last, and why they said yes: [paste 3 messy notes]
- The last campaign we actually ran, including what stalled: [paste]
- A page or competitor we keep losing to: [paste URL or copy]

Write a one-page campaign brief I can hand to a person on Monday.
Include:
1. The buyer and the moment they are in.
2. The single offer, in language I already use.
3. One channel for the next fortnight, and why not three.
4. The first asset to make (email, page, or post), with a rough outline.
5. Three claims we must not make, because I did not give you proof.
6. What I must still approve (send, publish, spend).

Do not invent results, prices, suburbs, or customer quotes. Do not recommend new software. Do not write the finished emails or posts. The brief is the deliverable.

Paste a messy version. The polite version will get you a polite brief nobody uses.

How do I use Claude for market research?

Research is the input to the brief. It is not a 40-page report.

Give Claude a real question a buyer would ask. Give it your current page, a competitor page, or a messy notes file. Ask it to list what a buyer can understand, what is missing, and what you should not claim.

Then check the claims yourself.

Claude can search and summarise. It cannot sit in your sales calls. It will sound sure about a suburb, a price, or a competitor feature it has inferred. If you cannot point to the source, cut the line.

Useful research outputs for a founder:

• A list of the questions customers already ask before they buy
• A comparison of how you describe the offer versus how a competitor does
• Gaps on your own page that a model (or a tired buyer) cannot resolve

Do not ask it to “do the market research.” Ask it to finish one artefact you will use in the brief this week. A one-page gap list becomes the brief. A trend deck becomes a tab you never reopen.

If the pile is a folder of transcripts or a spreadsheet, Chat starts to feel small. That is a file job. Stay here if the inputs still fit in the prompt.

How do I use Claude Cowork for marketing, and when is chat enough?

Chat is enough when the job fits in a conversation: a brief, a research memo, a first draft, a rewrite.

Claude Cowork is the file job. You hand Claude multi-step knowledge work across the files and tools you choose, then come back to something you can review. Anthropic lists Cowork on paid plans (Pro, Max, Team, Enterprise). Chat and Cowork share one home at claude.ai. Pick Cowork when you need a folder, a spreadsheet, a pile of transcripts, or a document back as a file.

Use Cowork when you have:

• A folder of call notes that should become one brief
• A CSV of last month’s enquiries that should become a shortlist
• A deck or doc that must come back as a file, not a chat paragraph

Stay in Chat when you are thinking on the page. If the job is a one-page brief, Chat is the calmer path. Give Cowork a narrow folder, not the whole shared drive. Anthropic’s own guidance is to stay close on work with real consequences: money, messages sent as you, important files.

Claude Code is a different product. It is Anthropic’s agentic coding tool, in the terminal, IDE, desktop app and browser. For marketing, open it only if the work is actually in a codebase. If you are not editing project files, you do not need it. Founders start in Chat.

How do I use Claude for social media marketing without publishing the first draft?

You use it as a drafter. You do not give it the login.

Paste the campaign brief, three posts you were happy with, and the one offer this fortnight. Ask for five options in your voice. Then delete the ones you would not say at a client lunch.

The first draft is raw material. It is not the post.

If Claude sounds like a press release, give it a sentence you have already used. If it invents a result, a suburb, or a customer quote, throw the line out. Five options. One channel. You pick one, edit it, and you tap publish.

How do I set up Claude for a repeatable marketing job?

Once the brief loop works once, make it boring.

In Chat, use a Project. Projects keep related work in one workspace with its own files, instructions and memory. Put the offer sentence, the audience, the words you never use, and last week’s brief in there. Next Monday you are not re-explaining the business.

In Cowork, add standing instructions so every session starts from the same rules. Anthropic also ships plugins that bundle skills, including a Marketing plugin. Skills and that plugin are optional. The repeatable job is the Project instructions plus the brief, not a slash command.

Write four lines and paste them at the top of the Project.

1. Who we sell to, in one sentence.
2. What we never claim.
3. What “finished” looks like (a one-page brief a second person can run).
4. What Claude must never do: send, publish, spend, or invent a proof point.

That is the setup. Not a stack. Scheduled tasks in Cowork can wait. Still a draft. Do not schedule a send.

Is Claude better than ChatGPT for marketing?

No useful answer starts with a winner.

The better tool is the one you will actually run a loop in. I use Claude when the job is a long document, a messy research pile, or a brief I want to argue with. I use ChatGPT when the team already lives there. The how to use ChatGPT for marketing page is that version of this loop.

If you bounce between both and finish nothing, you do not have a model problem. You have a loop problem.

Which Claude model is best for marketing work?

Start with Sonnet 5 for writing, analysis and a one-page brief. If you are not sure, start there.

Use Opus 5 when the research is long, the files conflict, or the brief needs deeper reasoning. Anthropic lists it for complex tasks and knowledge work, including docs and spreadsheets.

Use Fable 5 for long knowledge work that should run with fewer check-ins. Haiku 4.5 is the fast model for quick sorts and first cuts. A clear brief in Sonnet 5 beats a vague brief in a bigger model. I will not quote prices. Check Anthropic for the current list.

Should Claude send, publish, or spend on my behalf?

No.

Claude can assemble the work. You still remain the judge.

That includes Cowork, scheduled tasks, and anything connected to Gmail, a CRM or an ads account. Draft the reply. Do not let it send. Draft the post. Do not let it publish. Flag a wasted ad. Do not let it change the budget. If a customer cannot undo the action, a person taps the button.

My honest read

My honest read: Claude for marketing is a one-loop tool, not a department. Run research into a campaign brief in Chat this week. Use Cowork only if the job is a pile of files. Keep send, publish and spend in your own hands.

The guardrail is the same one I use everywhere. Narrow the access. Review before you widen it.

FAQ

What is Claude for marketing?

Claude for marketing is using Claude to finish one named job (research, a campaign brief, or a first draft) and then having a person decide what gets sent, published or spent. It is not a replacement CRM, and it is not a reason to install Claude Code on day one. Start in Chat at claude.ai with facts you already trust.

How do you use Claude for marketing?

Open Chat at claude.ai. Paste the facts you trust. Ask for a one-page campaign brief: audience, offer, one channel, one next action, and what must stay human. Edit anything you would not say to a customer. That is the first loop. Add Cowork later only if you need files.

Is Claude better than ChatGPT for marketing?

Neither wins in the abstract. Use Claude if you will actually run the brief loop there. Use ChatGPT if that is already the team’s habit. Switching models will not fix a missing process. I keep both pages on this site so you can pick the chat you will open on Monday.

What are Claude Cowork and Claude Code for marketers?

Claude Cowork is Anthropic’s file-and-task product for multi-step knowledge work on paid plans. Use it when the brief needs a folder, a spreadsheet or a document back as a file. Claude Code is the coding product. Marketers only need it if the work lives in a codebase.

Which Claude model is best for marketing?

Sonnet 5 is the default for everyday writing and analysis. Step up to Opus 5 (or Fable 5 for the longest jobs) when the research is messy or the brief is high-stakes. Use Haiku 4.5 for quick sorts. A clear brief beats a bigger model every time.

Should Claude send emails or publish posts for me?

No. Let it draft. You tap send, publish or spend. That rule holds for Chat, Cowork, scheduled tasks and any connector you attach. If the action cannot be undone by a customer, it stays human. Confidence in the draft is not a reason to skip the gate.

What to do this week

Pick one offer. Open Claude Chat. Paste the prompt above. Get a brief you can judge in thirty minutes.

Do not connect the inbox. Do not open Claude Code. Do not buy a plugin pack.

If you want a wider view of the system around this, the Digital Bond scorecard takes about four minutes and does not ask for an email.

If you would rather install the first loop with other founders, that is what AI Systems Lab is for. And if you want a person in the loop with you, I do that work as an AI consultant.

I will take a brief you can hand to someone on Monday. I will not take a stack you never open.