Measurement & business caseAgenticSeptember 12, 2026

The Earned Media Budget Defense

Suggested tool: Claude (it will tell you your sample is too small rather than building you a confident-sounding case on 18 data points)

The scenario

Budget season. Someone above you is going to ask why earned media deserves a line in the AI visibility budget when the owned-content vendor has a deck full of charts. You have industry studies that contradict each other, and none of them looked at your category, your buyers, or your brand. You need a number you generated yourself.

The prompt

Act as a finance partner to a communications team preparing a budget defense. I'm going to give you three things: a log of what AI models actually cited when asked buyer-stage questions about us, a list of our earned coverage from the last 12 months, and our current AI visibility spend by line item.

First, tell me what the citation log actually shows. What percentage of citations came from sources we own, sources we earned, and sources we neither own nor earned? Show your counts, not just percentages.

Second, cross-reference the earned coverage list against the citation log. Which specific placements are showing up in AI answers, which are invisible, and is there any pattern in outlet type, recency, or format that separates the two?

Third, tell me honestly where my sample is too small to support a claim. If I ran 18 prompts, say plainly what 18 prompts can and cannot prove, and what I would need to run to make it defensible.

Then draft a one-page budget recommendation with a specific allocation, the reasoning behind each line, and the single strongest counterargument someone in that meeting will raise. Write the counterargument as fairly as you can, then answer it.

Do not use any industry benchmark or third-party statistic unless I provide it. This case has to stand on my own data.

[PASTE CITATION LOG, COVERAGE LIST, AND CURRENT SPEND]

How to use it

Write down 15 to 20 questions a real buyer would ask an AI assistant when they're close to choosing. Not "what is [category]." More like "[your company] vs [competitor] for [use case]" and "how does [your company] handle [the thing procurement asks about]." These are the moments where money changes hands.

Run all of them across at least three models. Record every source cited in the answer: the URL, whether it's yours or someone else's, and what type of page it is. Owned site, product page, earned coverage, directory listing, forum, analyst note. This takes about 40 minutes and it is the only part nobody can do for you.

Pull your last 12 months of coverage into a simple list, and your current AI visibility spend broken out by line. Then paste all three inputs into Claude and run this:

Read the cross-reference section first. The interesting finding is almost never the headline percentage. It's discovering that the trade piece you were proudest of has never once surfaced, while a two-year-old byline you forgot about gets cited constantly.

Rerun the same prompt set next quarter. One measurement is an anecdote. Two is a trend line, and a trend line you generated yourself beats anybody's survey.

Why it works

You walk into the budget meeting with data from your own company, the only kind that survives a follow-up question.
It finds the gap between the coverage you're proud of and the coverage AI actually reads, and those two lists are never the same.
It gives you the counterargument before someone else does, so you're answering it instead of absorbing it.

Cowork version

For Claude Cowork, where the prompt can read the files in a folder.

Build an Earned Media Budget Defense from the files in this folder and save the outputs here.

1. Read the context file and write 20 buyer-stage prompts a real customer would ask an AI assistant when comparing us against the named competitors. Show me the list before you run it.

2. Run every prompt and capture each source cited: URL, domain, page type, and whether it is owned, earned, or neither. Build this into a structured citation log.

3. Read the coverage export and cross-reference it against the citation log. Identify which specific placements appear in AI answers and which never surface. Look for patterns in outlet, recency, format, and whether the piece names us in a headline.

4. Calculate the owned, earned and third-party split with raw counts alongside percentages. State plainly what this sample size can and cannot support, and what a defensible sample would require.

5. Draft a budget recommendation with a specific allocation, the reasoning per line, and the strongest counterargument stated fairly and then answered. Use only the data in this folder. Do not import industry benchmarks.

6. Save the citation log as AI Citation Log.xlsx with one row per citation, and save Earned Media Budget Defense.md as the one-page recommendation.

This prompt first ran in TIC. Read the original issue →