What MCP is, in plain words

MCP is a plug. On one side is the AI app you already use, like ChatGPT or Claude. On the other side is your ad data. Once they’re plugged together, you can ask the AI questions about your own account and it answers from your real numbers, not from guesses.

Pulse Pilot’s connector is read-only. The AI can look at everything, but it can’t change a bid, a budget or a keyword. If it suggests a change, you make it yourself, in Pulse Pilot.

We connected Claude to a real Pulse Pilot account, a vacuum-bag brand doing about $21,800 a month in total sales, and asked the four questions sellers ask most. The answers below are shortened, but the numbers are real.

Question 1: “How did last month go?”

The AI pulled the last 30 days and compared them with the 30 days before:

Its summary: spend dropped by a third, orders held, and ACoS fell from 37.8% to 23.1% because clicks were converting better. It also pointed out that TACoS (ad spend against all sales, organic included) was 6.7%, meaning ads were a small slice of the business.

The useful part: it flagged that the last 7 days would still rise, because Amazon credits a sale to the day of the click for up to a week. That’s the kind of detail people forget when they panic over a bad Tuesday.

Question 2: “Where am I wasting money?”

The AI looked for searches with at least 5 clicks and zero orders over 60 days. It found 11, adding up to $106, about 3% of the account’s $3,656 in spend.

Then it did something most search-term reports don’t: it split them into real waste and too early to call, using a simple test: expected orders = clicks × the product’s conversion rate.

The product converts about 20% of ad clicks. So 20 clicks “should” have brought about 4 orders. Zero out of four is a real signal. Twelve clicks should have brought about 2, and missing 2 orders happens by chance more often than you’d think.

(We wrote a whole post on this math: How many clicks before you cut a keyword.)

Question 3: “Is it my listing or my bids?” (the one that matters most)

If you take one thing from this post, take this one. It’s the most expensive mistake in Amazon PPC: cutting bids on a product whose real problem is the listing.

Here’s the principle. Amazon ranks products that sell. A click only turns into a sale, and a sale into organic rank, if your listing converts at least as well as the products next to it. If shoppers click your ad and then buy from the listing beside yours, no bid fixes that. You’re paying to send customers to your competitors.

So before you touch a bid, ask: does this product convert as well as the market?

The AI checked one 8-pack with a 66% ACoS. Its ad clicks converted at 11%. Pulse Pilot’s market benchmark for that product was 29.2%. The listing as a whole, counting every visitor with or without an ad, converted poorly too, at 18%.

Its answer: when both your ad clicks and your listing overall convert below the market, the problem is the listing, the price or the offer, not the bids. Lowering bids would cut spend and sales together, and ACoS would barely move.

Compare that with the account’s 6-pack, which converts at 27.6% against a market of 21.7%. That listing does its job, so for that product, bids and targeting are the right place to look.

How Pulse Pilot uses market conversion

The AI could answer this question because the market number is already in Pulse Pilot. It’s not a one-off report. It runs through the whole product:

• On your dashboard. Every product shows its conversion next to the market’s, with a plain verdict: at or above market, below market, or not enough data yet.

• In your own rules. In Pilot, market conversion is a condition you can build rules on, like any other metric. For example, raise bids only on keywords where you’re converting above the market, so your budget goes to clicks that can earn organic rank.

• In Autopilot. If you’d rather be hands-off, Autopilot uses the market benchmark too, so it isn’t judging your ads on ACoS alone.

The rule above is a balanced example to show how the condition works, not Pulse Pilot’s own algorithm. Autopilot and Copilot weigh more factors than any single rule.

Question 4: “Which searches deserve their own campaign?”

The AI sorted every search by orders and flagged the ones performing well but still sitting inside a broader research campaign:

Neither had a campaign of its own. The AI suggested moving each into a single-keyword exact campaign, so each one gets its own bid and budget instead of sharing them with dozens of weaker searches.

That’s not the only way to structure it. Here’s how the two options compare:

Pulse Pilot builds both. In Pilot, the Campaign Creator sets up a single-keyword campaign or a group campaign in a few clicks, and a harvest rule can move a search into its own campaign automatically once it proves itself. On Autopilot, a proven search moves into a research campaign first, then into its own single-keyword campaign.

What the AI can’t do

• It can’t change anything. That’s on purpose. It suggests; you decide.

• It can be wrong. The numbers come straight from your account, but the conclusions are the AI’s. Check anything important before you act.

• It only knows what’s in your account. It can’t see a competitor’s price change or a listing that got suppressed last night.

What Pulse Pilot’s connector can see

• Account, campaign, keyword, product-target and placement performance

• Search terms, with the campaign each one ran in

• Product sales including organic, sessions, and TACoS per product

• Pulse Pilot’s market benchmark for each product

• Rank tracking

• The change history: what changed, and when

• Your own rules

It also comes with ready-made prompts: find wasted spend, weekly account review, why did sales drop, diagnose high ACoS, and a product health check.

How to connect it (about 2 minutes)

1. In Pulse Pilot, go to Settings → Connect your AI and copy your connector link.

2. In Claude or ChatGPT, add it as a custom connector. Some AI apps require a paid plan for this.

3. Sign in to Pulse Pilot and approve. You can revoke access anytime from the same settings page.

Full steps: Connect your AI.