The Audit Checklist
If you want the short version, this is the whole thing. Work down it in order.
- Tracking — is the Insight Tag firing, are conversions defined, do they match the CRM?
- Metric layer — are you reporting landing page clicks or total clicks? Settle this before reading any other number.
- Audience settings — audience expansion off, LinkedIn Audience Network reviewed, exclusions in place?
- ICP alignment — what share of spend reached job titles you actually sell to?
- Funnel coverage — do you have TOFU, MOFU and BOFU, and what is the budget split?
- Retargeting depth — 180-day windows, segmented by engagement depth, or one lazy "all warm" audience?
- Format fit — is each format judged on its own correct metric?
- Creative health — frequency, fatigue, message variation, format mix.
- Destination — does every campaign have somewhere to send people, and is it a dedicated page?
- Pipeline — can you connect any of this to closed revenue?
Most published LinkedIn Ads audit checklists cover items 1, 3 and 8. Those are hygiene. The money is in 2, 4, 5 and 9, and they are the ones almost nobody checks because each requires looking at the account as a system rather than campaign by campaign.
Manual, Checklist, or Diagnostic?
"AI audit" is doing a lot of work in most marketing copy. There are three quite different things being sold under that name.
A summary. A model reads an export and writes a paragraph describing it. Genuinely useful for a status update, useless for finding problems, because a description of what happened is not a diagnosis.
A checklist run by a model. Fixed rules applied to your account: is audience expansion on, are you using exclusions, is frequency too high. Better. This catches real hygiene issues, and most accounts have several.
A structured diagnostic with benchmarks. The account is evaluated layer by layer, each layer scored, findings ranked by what would actually move pipeline, and every number compared against a stated sample of comparable accounts. This is what we mean by an audit.
The difference between the second and third is prioritisation. A checklist gives you fourteen things wrong with your account. A diagnostic tells you which three matter this month, and that is the entire value.
The Eight Layers
This is the diagnostic we run across managed accounts, and the same one exposed through the Kiin MCP server. The order matters — it runs top of funnel down, because a conversion problem is usually caused by something further up.
| # | Layer | What it checks |
|---|---|---|
| 1 | Thought Leader Ads | Presence, volume, author diversity, LP CTR, engagement rate |
| 2 | Conversation Ads | Open rate, click-to-open rate, CPL, send volume |
| 3 | Lead Gen Forms | Submit rate, CPL on submissions only, form friction |
| 4 | Single Image Ads | LP CTR, CPM, true LP CPC |
| 5 | Video Ads | View rate, completion at 25/50/75%, cost per completion |
| 6 | Full-Funnel Structure | TOFU/MOFU/BOFU coverage and budget split across them |
| 7 | GTM Alignment | ICP targeting quality, company targeting, waste |
| 8 | Content Strategy | Creative diversity, message variation, format mix |
Each layer produces a score and one prioritised action. The health score is a weighted average. The score itself is close to meaningless as a number — what matters is which layer is dragging it down, because that tells you where the money is going.
Layers 6 and 7 are highlighted deliberately. They are where most accounts fail, and they are the two nobody audits, because both require looking at the account as a system rather than campaign by campaign.
The Three Findings That Come Up Most
1. Budget going to people you do not sell to
Almost every account has this and almost nobody measures it. Targeting gets set up once, job titles drift, seniority filters are looser than intended, and audience expansion quietly widens everything. Six months later a meaningful share of spend is reaching titles that were never in the ICP.
It is invisible in Campaign Manager because demographic data lives on a separate screen from spend data, and nobody reconciles the two against the target list. Ask an assistant with account access "how much of last month's spend went to job titles outside our ICP" and you get an answer in seconds. It is usually uncomfortable.
2. No mid-funnel at all
The most common structural failure in B2B LinkedIn accounts: cold awareness campaigns and "book a demo" campaigns, and nothing in between.
Someone sees a thought leader ad, finds it interesting, does not click, and is never contacted again — or is immediately hit with a demo request they are nowhere near ready for. There is no 180-day retargeting layer catching engagers, video viewers, or people who opened a lead form and abandoned it.
This is layer 6, and it fails before layer 1 does in most audits. Teams obsess over creative while running no mid-funnel whatsoever, which is optimising the top of a funnel that has no middle.
3. Thought leader ads with nowhere to go
TL ads get set up because they work, generate strong engagement, and then send traffic to a homepage — or to nothing at all, because the objective was set to brand awareness and there is no destination.
The engagement looks great. The pipeline contribution is zero, so eventually someone concludes thought leader ads do not work and switches them off. What did not work was running the best awareness format on the platform without a conversion path behind it.
What an AI Audit Cannot Find
Being straight about the limits, because the category is being oversold.
Anything outside LinkedIn. If your CPL rose because someone changed the landing page, no ad audit will find it. The data does not exist in the account.
Whether your offer is any good. An audit can tell you a conversion rate is low. It cannot tell you the offer is weak, the pricing is wrong, or the market does not want the product.
Lead quality. Without a CRM connection, an audit sees leads, not whether any of them were real buyers. A campaign with a brilliant CPL and no closed revenue looks excellent right up until someone checks the pipeline.
Whether your creative is good. It can measure that one creative outperforms another. It cannot tell you why, and it cannot tell you what to make next. That is still a human judgement.
Context. A spike might be a conference, a competitor's funding announcement, or a viral post. The account shows the spike. It does not know what caused it.
The honest framing: an audit compresses several hours of exports and pivot tables into a two-minute conversation, and it is much more consistent than a human doing the same job at 6pm on a Friday. It is not a strategist.
How to Run One
Connect an MCP server to Claude or ChatGPT — the setup takes about five minutes and is covered step by step in connecting LinkedIn Ads to Claude. Then work through these in order:
- "Run a full audit on this account for the last 90 days."
- "Give me the three highest-impact fixes, ranked by likely pipeline effect."
- "For the top finding, show me the underlying numbers and which tool you used."
- "How does this account compare to benchmarks, and what is the sample behind those benchmarks?"
- "What would you check next that you have not looked at yet?"
Question three is the important one. An audit you cannot interrogate is a horoscope. If the assistant cannot show you which numbers produced a finding, do not act on it.
Question four separates real benchmarking from decoration. Several tools advertise benchmark comparison without stating how many accounts sit behind the number, over what period, or in which verticals. A benchmark without a stated sample is an opinion with a number attached.