Can Claude really do AEO for free?

For a point-in-time audit, yes. These four prompts walk Claude through mapping buyer questions, checking where you show up, diagnosing gaps, and drafting fixes. What it cannot do for free is run continuously. AI search answers change week to week, so a one-time audit goes stale. That is the part a platform automates.

Do I need to pay for anything?

No. The playbook runs on Claude and your own Google Search Console data, both free. An optional DataForSEO key adds competitor ranking data if you want it, but it is not required. You only pay if you decide to have the whole thing run automatically instead of by hand.

How is this different from the free scan?

The playbook is you running the audit yourself, prompt by prompt. The free scan is Loudmink running the same kind of audit for you in about ten minutes, no prompts and no account. Same goal, different amount of work on your end.

Why does a one-time audit go stale?

Our research tracking brands across every major engine shows only about 38% of citations survive from one week to the next. Engines re-search and re-rank constantly. Running the playbook once tells you where you stand today. Staying in the answers means doing it again, across every engine, every week.

Free DIY playbook

Do your own AI search audit with Claude.

Four copy-paste prompts. Find out what ChatGPT, Gemini, and Perplexity say about you, and fix what is missing. Runs on Claude and your own Search Console data. No account needed.

Start the playbook

New to how AI search picks brands? Read the guide first.

Before you start

What you'll need.

Claude

Claude Code or the Claude app. Free tier is enough to run the prompts.

Search Console

A Google Search Console property, connected via the open-source GSC MCP server.

Optional

A DataForSEO key for competitor ranking data. Skip it and the playbook still works.

Map

Find the questions AI answers about you

AI search engines break one buyer question into a tree of smaller ones, then answer those. Before you can see if you show up, you need the real list of questions a buyer asks on the way to choosing.

Prompt 1 · Map

I run [brand] at [domain], in the [category] space. My customers use AI search engines like ChatGPT and Perplexity to decide what to buy. Break the broad question "what is the best [category]" into the 15 to 20 specific sub-questions a real buyer would ask before choosing, including ones with constraints like price, location, use case, and "best for [segment]." Use my Google Search Console data to ground these in what people already search for. Output a numbered list I can paste into any AI search engine.

You get: A buyer-question list you can run against any engine.

Loudmink generates these from your site and tracks them across every engine, every day.

Check

See whether you show up

Run your questions and record the answer. Who gets named? What kind of source did the engine pull from: your site, a listicle, Reddit, a review page? This is where most brands learn they are missing.

Prompt 2 · Check

Here are my buyer questions: [paste Prompt 1]. For each one, tell me: would a current AI search engine likely recommend [brand]? Who shows up instead? What kind of source would the answer pull from (my site, a listicle, Reddit, YouTube, reviews)? Be skeptical: assume I am missing unless my site clearly answers that exact question. Output a table: question | shows me? (Y/N) | who wins | source type | why.

You get: A scorecard of where you show up and where a competitor wins.

Loudmink checks every question against the live engines, every day, and records who shows up instead.

Diagnose

Find the gaps that matter

Not every gap is worth fixing. Separate the ones where no strong page answers the question from the ones where the answer comes from a third-party source you are not part of. Then rank by intent.

Prompt 3 · Diagnose

Cross-reference the Prompt 2 table with my Google Search Console data. Split my gaps into two kinds: (1) content gaps, where no strong page answers the question, and (2) credibility gaps, where the answer comes from third-party sources I am not part of (Reddit threads, listicles, review sites). Rank them by buyer intent and how close I already am to winning. Output the top 10, each with the one specific reason I am losing it.

You get: A ranked top-10 gap list, content gaps and credibility gaps split out.

Loudmink ranks your gaps for you and flags which are worth fixing first.

Fix

Produce what closes the gap

Turn the top gaps into something you can publish: a page outline with the answer up top, an FAQ block, a comparison table, the schema to paste in, and the third-party places you need to show up.

Prompt 4 · Fix

Take the top 5 gaps from Prompt 3. For each, give me a ready-to-publish fix: a page or article outline with the direct answer up top, an FAQ block, a comparison table where it helps, and the JSON-LD schema to paste in. For credibility gaps, name the specific third-party places (subreddits, listicles, YouTube formats) where I need to show up, and draft the first one. Match my brand's tone from [domain].

You get: Draft pages, schema, and a first third-party post, ready to review.

Loudmink writes and publishes the fix, then keeps watching to see if it worked.

Track

Build a tracker and watch it over time

One audit is a snapshot. AI answers shift week to week, so the only way to know if your fixes worked is to check again. Have Claude set you up a simple log you re-run on a schedule.

Prompt 5 · Track

Set up a simple weekly tracker so I can watch this over time. Create a markdown or CSV log with one row per buyer question and a column for each AI search engine, where I record whether I showed up (yes/no) and who won, dated by week. Then give me a short routine I can re-run every week: re-ask the questions from Prompt 2, update the log, and flag any question where I dropped out since last week. Keep it copy-paste simple, no extra tools.

You get: A dated visibility log you update each week by re-running Prompt 2.

Loudmink keeps this log for you, every engine, every day, and charts the trend automatically.

Or run it with Loudmink

Rather not run five prompts? Loudmink does the whole playbook.

Connect Loudmink's MCP server once and Claude drives every step on your live data instead of an educated guess: it adds your buyer questions, checks each engine for real, surfaces who wins and where, scores your drafts against Loudmink's scorer, and tracks it all over time. One connection, no copy-paste.

Map · Check · Diagnose · Fix · Track, on real data

Terminal

claude mcp add loudmink npx loudmink-mcp --api-key=lm_xxx

Your API key lives in Settings, scoped to one product, and needs a Loudmink plan. See the MCP setup.

The honest part

These prompts work. Loudmink is the part that doesn't fit in a prompt.

Run the playbook once and you get a snapshot. But AI answers change every day, no one remembers every correction by hand, and copy-paste only reaches your own blog. Loudmink is a native AI agent built for the part that doesn't fit in a prompt: it runs all of this every day and learns your brand as it goes.

Learns as it goes

A native AI agent that gets to know your brand, your competitors, and what AI search engines say about you. It gets sharper the longer it runs.

Remembers what you tell it

Correct it once, like “we sell concrete tables, not just countertops,” and it keeps that fact for good, across every check it runs and every draft it writes.

Watches every day

Loudmink checks every engine, every day, and flags the moment you drop out of an answer. Not a manual pass you squeeze in once a week.

Runs the whole loop, and more

It maps the questions, audits each engine, finds the gaps, drafts and scores the content, and tracks the trend. Across your blog, Reddit, and YouTube, not just one page.

Free visibility report

Or skip the prompts. Let Loudmink run the playbook for you.

Free scan. No account, no prompts. See where you're missing in about 10 minutes.

Scan your brand free See pricing

Questions