MLT Marketing Lab is early software shown by invitation. What you're about to see — features, screens, pricing — isn't announced. Have a quick read and agree, and the site's yours.
Read the full terms · Not comfortable? Get in touch instead.
Four things in a loop, across every account you run and every channel it runs on. Everything else in the console exists to serve them.
We mark what's live and what we're still building, because you deserve to know which is which before you sign up.
The full pass, every time: search terms, negatives, ad copy, bidding and
budget, what's actually counted as a conversion, location targeting,
Quality Score, impression share lost to budget versus rank — and the
tracking parameters, because without utm_term your call
tracker reports every lead's keyword as “(none)” and months of data can
never be attributed to anything.
Every day, against each account's own baseline rather than a number somebody typed. Signal comes from the shop, the call tracker or analytics — the ad platform last, because it's the one source that keeps reporting success straight through a total collapse.
Every change arrives as a row: account, field, before, after, and why. Nothing touches a live account until you approve that specific row. Budget changes have no automatic path at any permission level, ever.
After a change it re-reads the account, explicitly checking the field it changed. An API call that returns success and a change that actually happened are different things — an empty field mask will cheerfully report the first while doing nothing at all.
The answer is split across the ad account, the analytics, the call log and the shop. So is the console.
Full audit loop, negatives, network splits, bidding, budgets, and ad copy edited in place rather than replaced — so an ad keeps its ID, history and learning.
Technical, on-page and content. Search Console, crawl, schema, page speed, robots.txt, internal linking.
The new one, and almost nobody is doing it properly. Its own section below.
Paid social alongside search, so budget can be judged across both instead of defended separately.
Connected to WordPress directly — fix the page, publish the post, correct the metadata. Not a ticket about the page.
Ads, landing pages, blogs and email, written as your agency in your client's voice.
Feed health, product performance, and order truth from the shop itself — the only number that settles it.
Campaigns, flows, segments, deliverability and sender authentication, which quietly stops mail sending when it lapses.
Prospect research, proposals and onboarding checklists built from what actually goes wrong on new accounts.
Your clients' customers are asking ChatGPT, Claude, Perplexity, Copilot and Google's own AI answers. Those systems pick who to name. Most agencies have no idea whether their clients get named, and no way to find out.
A fixed set of real customer questions, run every month against the major assistants. Is your client named? How are they described? Who gets named instead? Tracked over time, so you can show a client movement rather than opinion.
Audited and generated — Organization, LocalBusiness, Product, Service, FAQ, Article, Review. Structured data is how a machine understands a page without guessing, and guessing is how clients get described wrong.
llms.txt plannedThe emerging convention for telling models what a site is and which pages matter. Generated, kept current, and monitored — the same way a sitemap is.
Is robots.txt blocking GPTBot, ClaudeBot, PerplexityBot
or Google-Extended? Plenty of sites block them by accident and never
find out. Invisible, and expensive.
Written to be cited: clear claims, real numbers, direct answers to real questions. What gets quoted is different from what used to rank.
Which sources the assistants cite on your client's topics — so you know exactly which pages to get them mentioned on.
You connect each platform under your own credentials. Everything we read from, write to, or plan to.
Not a dashboard you have to interrogate. An ordered list of what costs money if it's ignored — with the accounts that are fine left quietly off it.
Delivered as an email you can forward, and as text when something can't wait until you check your inbox.
Calibrated per account against its own history, not a threshold someone guessed.
Ordered by consequence. Nothing on it that doesn't need a decision.
Blocked on a client's feed fix? It comes back when the blocker clears, not on a guessed date.
Draft the difficult email with the numbers already in front of it.
Every agency tool fails the same way — a wall of platform metrics answering no question anyone asked. A report should say what changed, what it cost or earned, and what you're doing about it.
Not with a chart of impressions. The client wants to know whether the money worked.
And the date it was measured. A number you can't trace is a number you can't defend on a call.
Ads, organic, email, calls and orders together, because that's how the business experiences it.
Your logo, your agency name, your voice. Not ours.
Claude Opus 5 drafts the narrative from the account's real numbers. You edit and sign it.
Generated on schedule and left for you. Nothing reaches a client unread.
Invite colleagues into your agency so they can see the same accounts, work the same watchlist, and pick up where you left off. Roles decide who can approve a change and who can only look.
Priced per account, never per seat — charging by the head just means fewer people see the alerts, which defeats the entire point.
Agency logo on client output · roles and permissions · per-plan token allowances · bring your own Anthropic API key · onboarding checklists per client · proposal generation from real prospect research.