Most “AI for PPC” posts show you a prompt and a polished screenshot of something that sounds smart. Then you try it on your own account, and the AI has never seen your data.

Optmyzr MCP fixes that part. It plugs Claude, ChatGPT, and other AI tools straight into your live Optmyzr account, so the answers come from your numbers instead of a guess.

For this post, we ran prompts through Claude with Optmyzr MCP connected, on Optmyzr’s own Google Ads account, and took screenshots of every answer.

If you haven’t connected Optmyzr MCP to your AI tool already, here’s how you can.

Note: To learn more about what an MCP is and what Optmyzr MCP offers, see the FAQs section below.

How to connect Optmyzr MCP to Claude or ChatGPT

Setting up Optmyzr MCP on Claude or ChatGPT only takes a few minutes, and you don’t need an API key or any code. You can sign in with your Optmyzr login, approve access, and your AI tool can read your accounts. Before you start, you’ll need an active Optmyzr account with your ad accounts connected.

Connect to Claude

1. Open Claude on the web or the desktop app.
2. Go to the Optmyzr connector in the Claude directory.
3. Click Connect.
4. Sign in with your Optmyzr login and approve access.
5. Start a new chat and make sure the Optmyzr connector is switched on in the tools menu.

If you’d rather add it by hand, open Settings, go to Connectors, and choose Add custom connector. Give it a name, paste https://tools.optmyzr.com/OptmyzrMcp as the server URL, click Add, then click Connect and sign in.

On a Claude Team or Enterprise plan, only an admin or owner can add the connector for the organization. After that, each person connects and enables it for themselves.

Connect to ChatGPT

1. Open ChatGPT and find the Optmyzr plugin in the ChatGPT plugin directory, or open the plugins section from the sidebar and search for Optmyzr.
2. Select Optmyzr and click Connect or Install.
3. When ChatGPT sends you to the authorization page, sign in to Optmyzr and click Allow access.
4. Open a new chat and enable Optmyzr as a tool.

To set it up manually, go to Settings, then Apps, and turn on Developer Mode. Choose Create, enter a name, paste https://tools.optmyzr.com/OptmyzrMcp as the server URL, set Authentication to OAuth, and click Create. If the Authenticate button doesn’t respond in the ChatGPT desktop app, connect with an API key instead.

Other AI tools

Claude Code, Microsoft Copilot Studio, Langdock, and other MCP clients connect with an API key instead. In Optmyzr, go to Settings, open MCP Integration, generate a key, and paste it into your client along with the server URL. Google Gemini also works through its Spark AI agent, which needs a Google AI Pro or Ultra subscription. The MCP user guide has the steps.

For the full walkthrough, see the Optmyzr MCP user guide.

Top 11 use cases of Optmyzr MCP (with real examples)

Once you have connected to Optmyzr MCP, you can combine any of these reads with your own questions. These 11 are a good place to start:

1. Run a health check in one prompt

If this is your first time using Optmyzr MCP, a health check is a good first read on the account. It compares your core metrics with the same period last year, so you can spot sudden performance shifts and flag broken tracking, and it ranks what to fix first.

You can use a simple prompt to get a read on your account’s health.

Prompt: Run a health check on my Google Ads account using Optmyzr. Compare the last 30 days to the same 30 days last year, then pull Optmyzr’s optimization opportunities and rank the top five by urgency.

The first table answers the question to settle before anything else: is the account getting more efficient, or is it just spending less? Everything else in the check reads differently depending on that answer.

Notice what Claude does next. It flags the metrics you shouldn’t quote yet. On a lead-gen account, return on ad spend can be misleading when conversion values change, and a jump in conversions deserves a check on what’s actually being counted. Your first next step is to confirm your conversion actions before you share any of it with a client.

Then it re-ranks Optmyzr’s opportunity list by what matters most, and points out items that should be decided together, like a budget fix and a bid target on the same campaign.

It also surfaces things you didn’t ask about, like a drop in Quality Score or automations that keep failing in the background. Those are easy to miss when you only look at performance.

What to do next:

• Fix the top-ranked item first, then re-run the check to see what moved.
• Verify your conversion tracking before you trust any efficiency numbers.
• Re-run the check weekly on busy accounts.

Try these prompts:

• Which of the opportunities you just ranked can I apply safely today, and which need a human decision first?
• Compare this month to last month at the campaign level and tell me which campaign changed the most.

2. Find budget-limited campaigns that still convert, then preview a rule

A campaign can look budget-limited, still beat its Target CPA, and not be a bid problem at all. This chained prompt finds the campaigns losing impressions to daily budget caps, then drafts a bid rule so you can see whether the target is really the constraint before anything goes live.

Prompt: Which of my campaigns are losing more than 10% of search impression share to budget but are still converting? Then, for the one that’s beating its Target CPA by the most, create a rule that lowers the Target CPA by 10% and preview it.

This gives you the shortlist that matters: campaigns that are performing well but held back by budget. That’s where extra spend is most likely to pay off.

Before it drafts the rule, Claude checks how each of those campaigns is actually bid. Only some had a target to adjust, so a target rule can only touch those. That check keeps you from building an automation that does nothing.

The rule comes back as a preview and nothing is applied. Claude also adds a few checks before you save it, like how the rule picks its campaign and whether the target is really the limiting factor.

That’s the useful part. You asked for a bid rule, and the answer pointed you back to the budget.

What to do next:

• Decide whether budget or bid target is the real constraint before you save any rule.
• If it’s budget, ask for a recommended increase, or a Rule Engine strategy that raises budgets only under conditions you set.
• Review the preview in Optmyzr before anything runs.

Try these prompts:

• Create a Rule Engine strategy that raises the daily budget for campaigns limited by budget and meeting their targets, with a maximum cap.
• Which of my campaigns are losing impression share to rank rather than budget, so more budget won’t help?

3. Find wasted search terms, without blocking your own brand

Pulling expensive zero-conversion queries takes a few clicks. Deciding which of them are safe to block is where most negative keyword lists go wrong. This prompt also protects valuable customer typos from getting accidentally blocked, so you cut real waste without choking off brand traffic.

Prompt: Using Optmyzr, show me the search terms that spent the most in the last 30 days with zero conversions on my Google Ads account. Then tell me which of them are misspellings of my own brand and which are genuinely unrelated, and which ones I should add as negatives.

Claude sorts the expensive terms, and the sorting is where the value is.

A lot of the “wasted” spend turns out to be people misspelling your own brand name. Those are customers trying to find you, so the right move is to leave them alone.

The part worth acting on is narrower: brand misspellings showing up in non-brand campaigns. Moving that traffic into your brand campaign gives you cleaner reporting and tighter control.

Claude also checks each candidate against your existing negative keywords before suggesting anything. Some clearly unrelated terms weren’t blocked anywhere, while others were blocked in some campaigns but still spending in another. That cross-check stops you from adding duplicates and missing gaps.

It also looks for the cause. Broad match on brand keywords is often what pulls in unrelated names, so tightening match types fixes the problem at the source instead of adding negatives forever. And zero-conversion brand queries are sometimes a landing page or tracking problem, not a negative keyword problem.

What to do next:

• Add the clearly unrelated terms as negatives, at the level where they actually spent.
• Keep brand misspellings in your brand campaign, using a shared brand list or extra brand keywords, instead of negating them.
• Review match types on your brand keywords.

Try these prompts:

• Stage negative keywords for the unrelated terms at campaign level and show me the preview before applying anything.
• Which of my brand keywords are on broad match, and what unrelated queries are they pulling in?

4. Find converting search terms, and find out which ones to trust

Converting search terms are the first place to look for new exact match keywords, but a conversion count alone doesn’t say which ones to trust. This prompt finds the search terms that converted at least twice in the last 90 days, checks which of them already have an exact match keyword, and also shows which campaigns have no negative keyword protection of their own.

Prompt: Using Optmyzr on my Google Ads account: (1) which search terms converted at least twice in the last 90 days, and which of those should I promote to exact match keywords? (2) Which of my campaigns have no negative keyword protection of their own? Don’t apply anything.

Claude cleans up the report first by removing duplicate rows, then checks whether each converting search term already has an exact match keyword. That keeps you from adding keywords you already have.

Here’s what makes this useful: it doesn’t promote everything that converts. One of the best-looking non-brand terms was held back, because the conversions behind it carried almost no value. That’s a sign a low-value action, like a page view or a small micro-step, is being counted as a conversion.

If you’d looked only at conversion counts, you’d have promoted the wrong term.

What to do next:

• Check which conversion actions are set as primary before you promote anything.
• Promote terms that convert with real value to exact match, so you control bids and ad copy.
• Look at the keyword behind each top converter. The search term can match exact while the keyword itself is broad.

Try these prompts:

• Break down my conversions by conversion action and tell me which ones carry little or no value.
• Which converting search terms are triggered by broad or phrase keywords, and what would exact match versions look like?

5. Audit which campaigns have no negative keywords

Campaigns without negative keywords usually show up as a problem after you switch them back on. This audit lists the campaigns that have no negatives of their own, so you can add them before a paused or experimental campaign serves again. This comes from the same prompt as use case 4.

Prompt: Which of my campaigns have no negative keyword protection of their own? Don’t apply anything.

The audit shows which campaigns have no negative keywords of their own. In this account, everything live was covered and the gaps were all in paused campaigns. That’s reassuring, and it’s also the check to run before you switch anything back on.

Claude goes further and points out which paused campaigns are most likely to come back, and which were running without negatives the last time they served. It also reminds you that account-level negatives can still apply, so “no negatives of their own” doesn’t mean nothing is blocking them.

What to do next

• Add starter negatives to any campaign you plan to re-enable.
• Consider a shared negative list for exclusions every campaign needs.
• Re-run the audit after every new campaign launch.

Try these prompts:

• Which of my paused campaigns would be missing negatives if I turned them back on today?
• Suggest a shared negative keyword list based on search terms that wasted spend across several campaigns.

6. Study last year’s peak, day by day

A monthly total hides which days did the work. Before you set a holiday budget, look at the daily shape of last year’s peak, then group it by weekday to see which days deserved more or less budget.

Prompt: Using Optmyzr on my Google Ads account, pull daily spend, clicks, conversions, cost per conversion, and conversion rate from Nov 20 to Dec 10, 2025. Chart it day by day, then group it by weekday and tell me which days I should have had more or less budget. Don’t apply anything.

Claude pulls the daily data and builds the chart right in the chat, so you can see the daily shape.

Then it groups the data by weekday, so patterns that hide in a daily chart become obvious.

The budget read is where it earns its keep: which days deserved more, which deserved less, and which you shouldn’t touch yet because the data is thin or skewed by a holiday.

It also refuses to over-read a single odd day. When one day looked like a delivery glitch rather than a real pattern, Optmyzr MCP said so and told us to investigate before changing anything. That’s the right instinct.

That’s the difference between a chart and analysis. (If you run a retail account, this is the exact prompt to use for your own Black Friday planning.)

What to do next:

• Shift budget from your weakest days to your strongest, in small steps.
• Investigate unusual days before you treat them as a pattern.
• Run this on your own peak period before you set budgets for the next one.

Try these prompts:

• Compare daily performance in the run-up to last year’s peak with the same stretch this year.
• Create an alert that flags when daily spend pacing is off during my peak window.

7. Benchmark your account against your industry

A metric only means something against a reference point. This prompt compares your account with other accounts in your industry vertical, then asks who your top auction competitors are over the last 30 days and which one is gaining on you fastest.

Prompt: Using Optmyzr on my account: how do I compare to other accounts in my industry vertical this month, and who are my top auction competitors in the last 30 days and which one is gaining on me the fastest?

Claude benchmarks your account against others in your industry vertical using percentile rankings instead of raw averages. That tells you where you’re genuinely strong and where you’re just average.

Then it reads past the rankings. It compares your trends with the industry’s, flags metrics that moved in a worrying direction, and points to structural differences like match type mix, bidding strategy mix, and how many campaigns you run. Those differences are often the real explanation behind the metrics.

What to do next:

• Pick one structural difference and test it, like adding more exact match coverage.
• Treat any suspicious metric as a tracking question first.
• Re-run the benchmark every quarter to see the trend.

Try this prompt: Show quarter-over-quarter and year-over-year benchmarks so I can see the trend.

8. See who you’re really competing against

This comes from the same prompt as use case 7. A new name in your auction can mean a new competitor or a match type problem. Auction Insights data shows which competitors overlap with you and how that overlap is trending, which helps you tell the two apart.

Prompt: Who are my top auction competitors in the last 30 days and which one is gaining on me the fastest?

You get a ranked list of the domains you compete with in the auction, how much each one overlaps with you, and whether that overlap is growing or shrinking. The biggest movers are usually the most interesting, and they’re often not who you’d expect.

Optmyzr MCP drills into the fastest-growing competitor and tells you where the overlap actually sits, which here was mostly on our own brand terms. It’s also clear about the limits: Auction Insights shows overlap, not spend or performance, so more overlap isn’t the same as winning more of the auction. And when overlap shows up on brand terms, your match types may be as much of the cause as the competitor.

What to do next:

• Drill into your top competitors and see which keywords you share.
• Decide whether brand protection or competitor campaigns make sense for you.
• Check your match types before you react to a new name in the list.

Try this prompt: Which competitors overlap most with my non-brand campaigns?

9. Check the alerts that fired and the ones you’ve configured

A long list of alerts is easier to act on once it’s ranked. This prompt asks which alerts fired recently and which ones matter most, lists the alerts you’ve configured and whether they’re on track, and pulls the last two weeks of change history to flag changes that could affect performance.

Prompt: Using Optmyzr on my account: what alerts have fired recently and which ones matter most? Then list the alerts I’ve configured and tell me which are on track. Also show the change history for the last two weeks and flag the changes that could affect performance.

Claude ranks the alerts by how much they matter. Budget overspend came first, then rising costs, then falling click-through rates, and it groups campaign-level alerts so you can see the pattern.

Then it lays out every alert you’ve configured in one table, showing what’s triggered, on track, or at target. In an account with many alerts, that replaces opening the alerts page and scanning it by eye. It’s also a quick way to spot gaps, like metrics you don’t have an alert on yet.

What to do next:

• Fix the loudest alert first, then check whether the others were symptoms of the same problem.
• Add alerts for the metrics you aren’t covering yet.

Try these prompts:

• Create an alert that notifies me when daily spend runs over my budget target by a set percentage.
• Which of my alert targets look unrealistic compared to the last 90 days of performance?

10. Line up a performance problem with the change that caused it

When spend jumps, the first question is whether something in the account changed. This came from the same prompt as use case 9: Claude read the alerts and the change history separately, then connected them, so you can check whether recent edits caused a cost spike or rule out routine campaign updates.

Prompt: Show the change history for the last two weeks and flag the changes that could affect performance.

Claude pulls the recent change history, groups what matters, and separates real changes from noise, like new campaigns that were created and paused within hours, or settings that were toggled and put back. Then it connects that history to the problem you asked about. Here, the log showed no budget edits on existing campaigns, so the overspend didn’t trace back to a budget change, and it suggested looking at cost by campaign next.

Change history works best as a way to rule causes in or out.

What to do next:

• Rule out changes first, then look at spend by campaign.
• Check who made a change and when before you assume it caused the problem.
• Ask for the change history before you ask why performance moved.

Try this prompt: Which changes in the last month could explain the shift in performance, ranked by likelihood?

11. Draft a brand new Search campaign from the chat

A first campaign structure takes time to build, and it’s easier to judge once it’s on the page. For Google Ads Search campaigns, Optmyzr MCP can draft a complete new campaign, including the budget, ad groups, keywords, negative keywords, and responsive search ads, as one change request. A person reviews it before anything goes live.

Prompt: Draft a new Google Ads Search campaign for [product or service] in [market]. Use a daily budget of [amount] and [bidding strategy]. Create ad groups by theme, add phrase and exact match keywords, add starter negative keywords, and write responsive search ads for each ad group. Put everything in one change request so I can review it, and don’t apply anything.

What comes back is a draft. It’s saved as a change request, checked against your account’s policies, and reviewed by a person. Nothing is created in your ad account until someone approves it. That makes it a fast way to get a first structure on the table, with your judgment still deciding what ships. Performance Max campaigns aren’t supported here: asset groups can be paused, renamed, or removed through MCP, but not created.

What to do next:

• Model the new campaign on your best-performing Search campaign instead of starting from a blank page.
• Read the keyword list and the negatives before you approve anything.
• Set a start date so the campaign goes live when you’re ready.

Try these prompts:

• Model a new campaign on my best-performing Search campaign, but for [new market].
• Write starter negative keywords for the new campaign based on search terms that wasted spend in similar campaigns.
• Write responsive search ad variations for each ad group and flag any headlines that repeat.

How to get better results from Optmyzr MCP

• Name the account and the date range. “Last 30 days” and “same window last year” are the two you’ll use most.
• Give it a job and a format. Be specific in your prompts. For instance, “Analyze my account” gets a book report. “Rank my campaigns by spend and show lost share to budget and to rank” gets a table you can act on.
• Ask the second question. The first answer is rarely the insight. Use case 3 only got interesting once we asked which terms were our own brand.
• Keep your judgment. MCP is fast at the analysis. The decision is still yours.

Ready to try Optmyzr MCP?

Pick the prompt for the task you keep putting off and run it on a real account today. Start with the health check, then ask the second question. If you’re new to Optmyzr, start a free trial. If you’re already a customer, connect Optmyzr MCP to Claude or ChatGPT.

FAQs

What is MCP?

MCP stands for Model Context Protocol. It’s an open standard that lets an AI assistant connect to outside tools and pull live data from them, instead of working only with whatever you paste into the chat.

Think of it as a plug. Without it, Claude only knows what you tell it. With it, Claude can go query a system, read a report, run a workflow, and then use what it found.

For PPC, that ends the export, upload, prompt, repeat loop. The connection stays open, and your AI pulls what it needs while you talk.

What is Optmyzr MCP?

Optmyzr MCP connects your AI tool to your Optmyzr account, so you can analyze and act on your ad accounts without opening the Optmyzr interface. The major advantage is that the numbers come from your live account instead of from the model’s guesses. Claude can still misread a result, so check anything you plan to act on. Changes you ask for are saved as change requests that a person reviews and applies.

It works with ad accounts connected to Optmyzr across Google Ads, Microsoft Ads, Amazon Ads, Meta Ads, LinkedIn Ads, and Yahoo! Japan Ads. Claude and ChatGPT connect with a sign-in. Claude Code, Microsoft Copilot Studio, Langdock, and other MCP clients connect with an API key. Setup steps are in the Optmyzr MCP user guide.

How is Optmyzr MCP different from other MCP servers?

The main difference is what comes back when your AI asks a question.

Google’s official Google Ads MCP server returns raw API responses. It’s read-only, and it needs developer credentials and a Python environment. Your AI has to interpret every result from scratch. (We wrote a full comparison of Optmyzr MCP and Google Ads MCP.)

Optmyzr MCP returns answers that already carry PPC logic: opportunities ranked by urgency, vertical percentiles, the alerts you set up, and Rule Engine previews. It can also draft changes, with a person approving each one.

Does Optmyzr MCP change my account on its own?

No. Changes made through MCP are saved as a draft change request. Nothing reaches your ad account until a person reviews and applies it.

Which AI tools work with Optmyzr MCP?

Claude and ChatGPT connect with a sign-in. Claude Code, Microsoft Copilot Studio, Langdock, and other MCP-compatible tools connect with an API key. Google Gemini connects through its Spark AI agent.

Which ad platforms does it support?

Google Ads, Microsoft Ads, Amazon Ads, Meta Ads, LinkedIn Ads, and Yahoo! Japan Ads accounts connected to Optmyzr. Some reads, like Auction Insights and industry benchmarks, are described for Google Ads accounts.

Can Optmyzr MCP make changes to my account?

It can draft them. Changes are saved as a change request that a person reviews, and nothing reaches your ad account until it’s applied.

What’s the difference between Optmyzr MCP and the Google Ads MCP?

The Google Ads MCP is read-only and returns raw API data. Optmyzr MCP returns analysis with PPC logic built in, works across more platforms, and can draft changes for human approval.

What’s the difference between Optmyzr MCP and Sidekick?

Sidekick works inside Optmyzr and knows which page you’re on. MCP brings Optmyzr data into your own AI tools, where you can combine it with other data and tools.

Do I need to know how to code?

No. For Claude and ChatGPT, you add the connector and sign in.

How much does it cost?

Optmyzr offers a free trial. For details on which plans include MCP, check your Optmyzr account or contact the team.