Emilia Korczynska is Head of Marketing and co-founder at ZenABM. Before that she was VP of Marketing at Userpilot for nearly 7 years, where she ran a large ABM program primarily on LinkedIn ads.

You can audit your LinkedIn ads with AI in minutes instead of a workday, as long as the AI can read your account and you keep the decisions.
This is the six-area checklist I would run, with the prompts, the settings most audits miss and the order to fix things in.
The easiest way to get started is to give Claude access to your account through ZenABM's MCP server, and use expert skills to audit your account data.
Here is the whole audit on one page:

Run these six areas in this order, because tracking breaks everything below it. Patrick Cumming sequences it the same way: tracking, audiences, budget, then creative. Most audits start with creative, which is backwards.
Area | What it catches | What the prompt does |
|---|---|---|
1. Tracking and conversion setup | Conversions that stopped firing | Lists every conversion and when it last fired |
2. Account structure and naming | Orphan campaigns still spending | Maps campaigns by objective and flags stale ones |
3. Audience and targeting | Non-ICP titles and overlapping audiences | Lists audience sizes, off-ICP spend and impression hogs |
4. Creative and format performance | Decaying ads | Ranks ads within each format by landing page CTR |
5. Budget and wasted spend | Spend with no engagement | Shows spend against landing page CTR trend |
6. Attribution and pipeline | Campaigns that touch deals | Joins ad touchpoints to open and closed-won deals |
Free resources: the /linkedin-abm-audit skill runs the audit for you. The ABM Benchmarks Report (2026 LinkedIn ABM benchmarks across 161,256 ads and 211 companies) gives you medians to grade against. The LinkedIn Ads Count Calculator tells you how many ads your monthly budget can actually feed.
AI replaces the boring 80% of an audit: pulling in the data, the sorting, the cross-referencing and the first pass at spotting what looks off. What stays human is deciding whether a finding matters for your business, your ICP and your stage of growth, and where the money should go next.
Gabriel Ehrlich, who runs the LinkedIn ads agency Remotion, shows what the mechanical part finds when someone finally looks. In his audit post, the LinkedIn Audience Network was eating 62.5% of a €28K monthly budget, and switching it off takes about four seconds.

A good AI audit gives you a clean, sorted picture of reality and flags anomalies. It should not decide budget allocation, and it should never change a campaign without your explicit approval.
Connecting AI to live data changes two things compared with working from Campaign Manager exports: completeness and metrics.
The first is completeness. Instead of auditing the ten campaigns you had time to export, the AI reads every campaign, ad set and ad in one pass and ranks them, so nothing hides in the long tail.
The second is the metrics. Campaign Manager's CTR counts every click on the ad, including "see more" clicks and image clicks, so it flatters ads that earn curiosity clicks. Landing page CTR is the cleaner measure: landing page CTR = landing page clicks ÷ impressions. For example, 40 landing page clicks on 10,000 impressions is a 0.4% landing page CTR. Landing page CPC works the same way: ad spend divided by landing page clicks. (In ZenABM this is called eCTR, our effective CTR, and it is the same thing as landing page CTR.)
Campaign Manager does not show either one by ad without an export and manual division. ZenABM's dashboards show landing page CTR and cost per landing page click out of the box. They also tie spend to your CRM, so the audit ends in pipeline rather than vanity engagement.
ZenABM also pulls first-party intent signals directly from the LinkedIn API, so account-level engagement comes from the same source LinkedIn itself uses. Here is ZenABM's company engagement table, with each company's ABM stage and engagement:

There is no official LinkedIn connector for Claude or ChatGPT, so the connection goes through a third-party MCP server that reads the LinkedIn Ads API and joins it to your CRM. With ZenABM, it takes four steps:
ZenABM's MCP server writes as well as reads: you can create and manage campaigns, ad sets and ads straight from Claude, and every change is confirmed first. This recording shows Claude building LinkedIn ads end to end through the server:

The full connection process and the broader reporting workflow are in the companion guide on analyzing LinkedIn Ads with Claude Code.
The six areas are in the table at the top of this post. Treat the prompts below as starting points and adjust the thresholds to your account. Here is the kind of clean, sorted picture the AI hands back when it runs an audit skill:

An audit that skips tracking audits numbers it cannot trust. Philip Ilic calls this "Layer 0": half the time the data cannot be trusted because tags are broken, UTMs are missing and the CRM is a mess.
Confirm that the Insight Tag fires, conversions are defined correctly, UTMs are consistent and the CRM sync is intact. The AI can cross-check that the conversions in the account match the events landing in the CRM.
Check my conversion tracking: list every conversion action, whether it has fired in the last 14 days, and any campaign optimizing for a conversion that has recorded zero events this month.
Here is where you check whether your Insight Tag is active in Campaign Manager:

Look for the silent break. A conversion that stopped firing three weeks ago will not announce itself, but every campaign optimizing toward it has been wasting budget ever since. Grade each event with a simple scale:

Messy structure hides waste. The AI should map the account (campaigns by objective, ad sets by audience) and list any legacy campaigns still spending that nobody remembers launching. In LinkedIn's current naming, a campaign (formerly a campaign group) holds ad sets (formerly campaigns), and ad sets hold ads.
Map my account structure. Group campaigns by objective and ad format, flag any campaign with spend in the last 30 days that has not been edited in 90, and list naming inconsistencies.

I have lived this one. In my post "The last two nights I was dreaming in spreadsheets", I described our own account with 1,382 active ads, where CTR exports alone said nothing about what worked. After an audit call with Gabriel Ehrlich, who told me I tend to overcomplicate things, we restructured from 17 simultaneous ABM campaigns down to 3 to 5.
The orphans are the tell. A campaign still spending months after its last edit is usually a test someone forgot to turn off, and it is the easiest budget to reclaim. If the AI cannot group campaigns cleanly because the names follow no pattern, that is a finding in itself.
Check who actually saw your ads, not who you targeted. LinkedIn delivers to more than what you pick, and Campaign Manager's Demographics tab shows real delivery, though only the top results. That is why Anthony Blatner pulls the full list of companies and job titles through the API.
Cristina Maliţa's 30-minute audit of a B2B SaaS account found 35 to 40% of budget going to people with nothing to do with the ICP, because targeting was too broad and there were no exclusions. After narrowing targeting and adding exclusions, cost per lead fell from $200 to $90 and cost per booked call from $400 to $250 within 10 days, with the same budget and ads.
Also check audience size against your budget. Experts diverge on the number: Philip Ilic keeps audiences under roughly 100,000, Marc Woodland uses 50,000 to 150,000 and Charlie Mora-Smith 20,000 to 80,000. The principle they share is to match audience size to budget. LinkedIn's own floor is 300 members per ad set.
Audit my targeting: list ad sets with audiences under 30,000 or over 100,000, any non-ICP job titles I spent over $200 on this month, and the companies eating the most impressions without engaging.
The answer matters most in two places. The first is overlap: when two ad sets target nearly the same audience, the account pays to compete with itself. The second is ICP drift, where budget leaks to titles outside your ICP because no one revisited a targeting setting.
I also keep a match-rate warning in mind. In a post from March 2026, I shared an audience pushed from HubSpot that showed a match rate above 90%, while only about 33% of the HubSpot list's companies were actually in the audience. Always compare the audience's real company count with your list.
Company engagements in ZenABM shows which specific accounts absorb impressions, how they engage and where they sit in the ABM funnel, so the targeting audit rests on named accounts instead of aggregated audience stats. Here is the company engagement view:

The MCP version of the same check sorts accounts into warm, cold, budget-hog and bad-fit buckets. This recording shows it running:

To check job titles, the persona audit skill sorts title delivery into CORE, SECONDARY and out-of-ICP, as in the persona audit recording earlier. ZenABM's Job title engagement then lets you exclude titles that should never have been served.
Tim Davidson reported another targeting miss in a post on a LinkedIn rep-managed account. It ran with audiences of 600,000 to 1.1 million, maximum delivery bidding and the Audience Network on, and his fix was manual CPC bidding, audiences cut to 100,000 to 400,000 and remarketing added.
Also watch for a LinkedIn bug I found in 2026. Edits to a saved audience do not reach ad sets already using it until you reselect that saved audience in each ad set. So after you change a saved audience, check that the ad sets actually picked it up.
Most audits get sloppy here by comparing incompatible formats. A Thought Leader Ad and a single image ad play different games, so benchmark each format against itself.
For context, the ABM Benchmarks Report shows Thought Leader Ads at a 2.68% median CTR and $2.29 median CPC, against 0.42% and $13.23 for single image ads. Those figures count clicks differently, so they are not like-for-like. Compare formats on landing page CTR and landing page CPC instead. ZenABM's Zena also checks your numbers against anonymised benchmarks from other workspaces, so you can see whether a landing page CTR is good for its format. You can also run the whole audit from Zena's chat with the LinkedIn ABM audit skill:
Rank ads within each format, then flag the ones in decline:
For each ad format separately, rank ads by landing page CTR and landing page CPC. Then give me a decaying-ads report: ads whose landing page CTR has fallen two or more weeks running above 1,000 impressions.

Read the decaying-ads list first. Creative fatigue on LinkedIn is gradual, so an ad slides for weeks before a monthly dashboard shows it, and by then it has already wasted the spend. A two-week slide in landing page CTR means you refresh or pause the ad before it costs another month. ZenABM's Bidding agent flags ads whose click-through rate is sliding and pauses them once you approve.
This audit table shows the typical fatigue patterns the AI surfaces, such as ads that peak in one week and fade:

For Thought Leader Ads, also break out performance by author, because the same message often performs very differently depending on whose face and voice carries it. The Thought Leader Ads guide covers how to pick and brief the authors.
This is the section that pays for the audit. The AI should show spend concentration by format, audience and funnel stage, and surface ad sets with rising spend and falling engagement. It should also name the accounts hogging impressions without ever clicking through. Each of those is a budget line you can reclaim.
Show me wasted spend: ad sets with the most spend and lowest landing page CTR, ad sets with rising spend and falling engagement over 4 weeks, and the 10 companies with the most impressions and least engagement.
Here is a waste-ranking table of the kind this prompt produces, ordering ad sets by spend against landing page CTR:

Too many ads for the budget is the other classic leak. Ali Yildirim sees 20+ active ads on a $100 daily budget, which gives each ad $4 a day, so nothing exits learning. His fix is 3 to 5 ads per campaign, and the LinkedIn Ads Count Calculator does the maths for your budget.
Check delivery against your plan too. In my ZenABM diary posts, our Document ads underdelivered at 6,913 impressions against a 15,700 target (44%), at a $42 CPM against the $26 target. I only caught that delivery miss because I compared plan to actual.
ZenABM's engagement score tracks current against total engagement per account, so the waste signal arrives with account-level context. The Companies table below shows impressions, clicks and engagements per company, which is how impression hogs show up:

A B2B audit that ends at lead counts has not finished. The real question is which campaigns touched the deals that opened and the deals that closed.
Which campaigns and ad sets were the most common touchpoints before open deals and before closed-won? Which spend produced no pipeline at all this quarter?

The answer reframes the audit. A campaign with a mediocre CTR that precedes half your closed deals should be defended, not cut, while a campaign with a great CTR that never appears before a deal is the opposite. Without the CRM join the two look the same, which is why audits that stop at engagement so often recommend cutting the wrong thing.
ZenABM's analytics dashboards handle that join natively, connecting LinkedIn Ads data to deal records in HubSpot, Salesforce or Attio. The deal attribution dashboard shows the touchpoints behind each deal. This recording shows deal attribution by campaign over time:

For the influenced-pipeline view, this is ad spend next to pipeline and revenue:

The six areas cover performance, but a surprising amount of waste hides one level down in campaign settings. Here is the one-glance version:
Setting | Wrong default | Right choice | Who recommends it |
|---|---|---|---|
Bidding | Maximum delivery | Manual bidding, starting at or below the suggested bid | Marc Woodland, Tim Davidson |
Audience Network | On | Off | Gabriel Ehrlich (our webinars) |
Audience Expansion | On | Off | Tim Davidson, AJ Wilcox |
Location | Recent or permanent | Permanent | Marc Woodland, Anthony Blatner |
Optimize for high value clicks | On | Off | Our 30 Sep 2026 webinar |
Campaign creation | Accelerate | Classic editor | Our 17 Feb 2026 webinar |
LinkedIn offers maximum delivery, cost cap and manual bidding, and the right choice depends on the objective and format. Maximum delivery spends the daily budget as fast as it can, which inflates CPMs. In our LinkedIn ad campaign setup webinar (17 Feb 2026), we recommended avoiding it and not bidding above the recommended bid at the start. Otherwise you are likely to overpay and drive up the cost of the whole auction.
Manual bidding also depends on what the bid charges for, so a bid set against the wrong charging model inflates cost without anyone noticing. The audit should put each campaign's bid next to its objective and flag mismatches.
For every active campaign, list: objective, bid strategy (maximum delivery, cost cap, or manual), the bid amount, and what that bid charges for (impression, landing page click, engagement, send, or video view). Then flag any manual bid priced more than 20% above the median for that ad format, and any bid strategy that does not fit the campaign objective.
ZenABM's Bidding agent checks every bid change against LinkedIn's suggested bid before you commit, and changes budgets in bulk.
Review daily and lifetime budgets, whether spend paces evenly or front-loads, and whether any campaign is too small for its audience. The classic finding is a campaign with no end date that has been spending for months past the moment it was relevant.
Show me, per campaign: daily and lifetime budget, how spend paced across the last 30 days (even, front-loaded, or under-delivering), any campaign whose budget is too small to serve its audience size, and any campaign with no end date that has spent for more than 60 days.

These toggles are the budget leaks that go unseen. Audience Expansion lets LinkedIn serve ads to profiles it considers similar to your audience, which dilutes ABM targeting. The LinkedIn Audience Network extends delivery to third-party apps and sites, with the same targeting, bid and budget. In our webinar on planning ABM strategy with AI (30 Sep 2026), we called these the two settings you always disable. Gabriel Ehrlich's numbers show why: LAN traffic showed about a 3% CTR when his real LinkedIn CTR was around 0.4%.
You can see the split in Campaign Manager under Breakdown, then On/Off Network, which separates Audience Network results from On LinkedIn results.
List every campaign with Audience Expansion enabled and every campaign delivering on the LinkedIn Audience Network. For each, show the spend, impressions, and engagement attributed to the expanded or off-network delivery, and compare account penetration on the core audience versus the expanded one so I can decide what to switch off.
Here is ChatGPT doing that check, listing campaigns with Audience Expansion and Audience Network delivery:

Location is the third default that goes unnoticed. LinkedIn serves "recent or permanent" locations unless you change it, so B2B campaigns should be set to permanent. Marc Woodland lists it among his audit mistakes, and Anthony Blatner found an enterprise account with 5 campaigns on the wrong setting. Also keep "optimize for high value clicks" off, because LinkedIn is notoriously bad at estimating what a high-value click is.
ZenABM's Budget Savers keeps Audience Network and Audience Expansion off on every ad set, including ones created later, and re-checks hourly.

Frequency capping in Campaign Manager is only available on the brand awareness objective, with a seven-day window, so check that setting for any awareness campaign. Across every campaign, the deeper check is alignment: the objective, the optimization goal and the bid should all point at the same outcome. A brand awareness campaign optimizing for reach but judged on landing page clicks is being graded against a goal it was never set up to reach.
For each campaign, put the objective, optimization goal, and bid side by side and flag any case where they do not point at the same outcome. For brand awareness campaigns, show the frequency cap setting and the accounts currently seeing the highest weekly frequency.
For accounts hogging impressions, ZenABM shows which companies soak up impressions without clicking, and you can exclude them in a click with Exclusions. That caps how much budget any one company absorbs.
An audit is only as good as the actions it triggers. Work the findings in this order, because it puts money back to work fastest:

For step 5, ZenABM flags accounts by engagement score and ABM funnel stage, so the handoff happens while the timing still holds. Here is how stages are defined:

The engagement score shows who is heating up:

And this is the same data pushed to company records in HubSpot:

None of this requires the AI to touch your account, because it proposes and the team decides. For any change in your account, such as pausing an ad, the MCP server asks for your permission first, so it never makes changes on its own.
I might be a bit biased as a co-founder, but I also know first-hand what it is like to struggle with Campaign Manager's hidden settings. Here is which ZenABM capability does which step of the audit:

One G2 reviewer, a VP of Marketing, put it this way: "ZenABM allows me to quickly query my LinkedIn ad performance data with AI," and ZenABM is rated 5.0/5 from 6 reviews on G2. Plans start at $59/month billed annually.

If you would rather not paste a dozen prompts, one command can run the whole audit. ZenABM's /linkedin-abm-audit skill grades your live account and returns a fix list ordered by dollars at stake. It is part of the free Claude skills at ZENABM/linkedin-abm-skills, and you can read the audit skill before you install it.

The skill produces a scorecard, checks the ad count against budget, grades each format against benchmark medians, lists decaying ads and impression hogs, and returns a prioritized fix list as branded HTML and PDF. In Claude Code, install it with two commands:
/plugin marketplace add ZENABM/linkedin-abm-skills
/plugin install linkedin-abm-skills@zenabmOn Claude Desktop or web, download the zips from the releases page and upload them under Customize, then Skills.

If you want a second opinion, Claude Ads is a free, open-source Claude Code skill by Daniel Agrici (AgriciDaniel/claude-ads). It runs 250+ automated checks across ad platforms and gives a weighted health score, working from pasted data or exported Campaign Manager CSVs.

Run the audit this month, fix the three biggest leaks, and then ask yourself why you waited a quarter last time. Start a free trial and run the first one this week, or book a demo if you want to see it first.
Yes. With your account connected through an MCP server, Claude, ChatGPT or Cursor can run the mechanical part of a LinkedIn ads audit in minutes: pulling every campaign, ranking by landing page CTR and CPC, detecting decaying ads and finding wasted spend. Strategic calls about what to change stay with you.
Yes. ZenABM's open-source /linkedin-abm-audit skill scores your account, grades each format against benchmarks and returns a prioritized fix list. Claude Ads is another free option that works from exported Campaign Manager CSVs. Connecting live LinkedIn and CRM data needs a ZenABM account, which starts with a free trial.
It checks six areas: conversion tracking, account structure and naming, audience and targeting, creative and format performance, budget and wasted spend, and attribution to pipeline. The connector adds metrics Campaign Manager does not show, such as landing page CTR and cost per landing page click.
Turn off the LinkedIn Audience Network, Audience Expansion and maximum delivery bidding, and set location to permanent instead of recent or permanent. Also switch off optimizing for high value clicks. These defaults serve LinkedIn's spend, not your pipeline, so keep them off, because new ad sets can ship with them on again.
Yes. Every audit check is a read, so nothing changes. If you later ask the AI to act on a finding, such as pausing an ad, the MCP server asks for your permission first and never makes changes on its own.
Run it monthly once it is automated. LinkedIn ad performance decays gradually, so a quarterly audit catches waste late. Saving the audit as a repeatable prompt or skill lets it run on the first of every month, with findings you can compare over time.
About the author