At 9:12 on a composite Tuesday, I’m at my desk looking at an account whose CPA has doubled since the night before. Same budgets. Same bids. Same search terms, more or less.

I know this scene because a client once swapped a landing page without telling me, and my Quality Score tanked overnight. The Tuesday that follows stitches together the kind of account work that taught me where to look when the numbers went bad. The first move was to stop looking at the ad account.

Tuesday: the account is clean, but the page isn’t

Change history is clean. Search terms are the usual mix. Budgets are pacing. In the keyword view, expected CTR and ad relevance look fine, but landing page experience has slipped to below average across the core exact terms. Google includes landing page experience in Quality Score for a reason. The ad is only half the promise. The page has to keep it.

I click through the ad. The old page put the offer in the headline, the price range above the fold, and the phone number in the header. The new one opens with a full-width hero image and a headline about craftsmanship. The form I’m paying $40 a click to feed is somewhere near the footer. It looks like a page someone was pleased to approve.

I run the same checks I’d run before calling anyone:

  1. Check change history for an account edit.
  2. Check auction insights for a new competitor.
  3. Check the feed if it’s Shopping.
  4. Click the actual ad on a phone, like a customer would.

Step four takes eleven seconds. The URL is the same slug with a new template ID on the end. The headline I wrote the ad to match word for word is gone. The page still loads fast; it just no longer says the thing I’m buying clicks to promise.

I’ve seen that failure in other accounts. One practitioner write-up put it cleanly: a cleaner redesign cut conversion rate by 35% because the clutter that got removed was information visitors needed to say yes. Google doesn’t grade design taste. Its landing page experience scale runs above average, average, and below average. A page can look better to the people who made it and work worse for the people who searched.

1 a.m.: two pages explain what the bids couldn’t

Later, I put the old page beside the new one and read both like someone who has just searched the keyword. The old headline names the service, the city, and the turnaround time. The new headline names none of them. The old page gives a price qualifier in its first 100 words. The new page spends 400 words on the company story before it mentions what the company sells.

I used to tell clients that copy like that could convert if the design was clean. I was wrong about how much a clean design could make up for a missing answer. The page has to tell a visitor, quickly, whether the ad brought them to the right place.

By 1 a.m., the chain is in my notebook for the call. Impression share hasn’t collapsed. Click-through rate has barely moved. Conversion rate has fallen, and average CPC has risen as the landing page grade weighs on ad rank. Say you’re spending $20k a month. That combination can turn a $68 CPA into a $140 CPA without anyone touching a bid. The account can be unchanged while the cost of what it delivers changes completely.

The cruel part is how reasonable the new page looks: faster, prettier, clean on mobile. What the deck calls a refresh, I call removing the sentences that did the selling. Another practitioner described a similar break when a client changed shop features and made preset filters obsolete. Conversion rate dropped although the ads and bids hadn’t changed. Category and filter pages are fragile PPC landers that way. One checkbox upstream, and the ad promises something the page no longer delivers.

9:40 a.m.: the people with the login get the call

I call with screenshots. In this version of the conversation, the client is cheerful about the refresh. The old page felt dated. The site agency said the new template would load faster and rank better. Nobody thought to tell the ads person because, in their minds, the ads and the website are two separate jobs.

I look at the notebook: $68 turning into $140. I try to explain that removing a headline and moving a form can change the economics of traffic we’re already buying. It’s awkward to be the person objecting to a prettier website. It’s worse to wait while the prettier website spends the budget.

The client offers to change it back if I think it matters. That is the part I keep hearing. The person buying the clicks doesn’t have edit access to the page the clicks land on. The people with edit access don’t know which promise brought those visitors there. In the composite, the headline goes back that afternoon and the numbers recover over the next two weeks. The access problem remains. I’m accountable for revenue that depends on a page I can’t touch.

Desk with old and new landing pages open side by side

The SaaS startup makes Tuesday a recurring appointment

One swapped page taught me where to look. A SaaS startup whose offer changed every quarter taught me how often I might have to look there. Each change gave the ads and the site another chance to disagree: free trial in one place, demo-only in another, usage-based pricing on a page that hadn’t caught up. An ad written to match last quarter’s offer can keep running long after everyone involved thinks that offer is gone.

The sensible routine was unglamorous. Every Monday: crawl the pricing and features pages, screenshot the hero copy, and compare it with ad copy and sitelinks. Then check the smaller places a promise can hide: an FAQ, a comparison page, an ad customizer. That catches a lot. It doesn’t make the underlying arrangement any less strange. A change to the product offer has to find its way through several people’s separate workflows before the account says the right thing again.

I used to think the match type structures I built by hand protected accounts from this kind of mess. Exact terms split by intent felt disciplined, and sometimes they were. But more ad groups also meant more headlines to keep in sync with a changing site. Tight structure didn’t give me control of the offer. It multiplied the places where the old one could survive.

Picture the quarter when a startup kills its free trial on Tuesday and replaces it with book a demo. The trial is gone from the main pricing page but still appears in the FAQ, two comparison pages, and the ad customizers. For ten days, the account pays for people who want something sales can no longer give them. Bids optimize toward form fills while the sales team opens calls by explaining the mismatch. The account can look efficient. The business is paying for apologies.

That’s the memory I bring to an agency demo of a white-label AI SEO dashboard. The tracked-visibility chart may look good, but I want to know who changes the page when the offer changes. Price math can be vague, too: one reseller lists $199.99 a month for 50 articles up to $499.99 for 200, while bigger names stay quote-based. Those details matter to an agency’s margin. Neither one gives the agency control of what a buyer reads.

The old offer gets a second life in AI answers

When an agency owner asks me which white-label AI SEO product to buy for AI visibility retainers, I think of that Tuesday. Picture the same SaaS client now. The offer changes, the site lags, and the old pricing page is available to be quoted in ChatGPT or Google’s AI Overviews while a buyer compares options. Content designers at the UK Department for Business and Trade have described stale GOV.UK pages feeding Google AI Overviews old information, leaving teams to chase the pages that keep surfacing. The problem is familiar: the public answer can outlive the offer behind it.

AEO troubleshooters describe the repair in ordinary website terms. An engine may keep quoting an outdated page even when Google-Extended is blocked and the page still appears in AI Overviews. The fix they cite is to get the updated page recrawled in indexable HTML and allow about 24 hours. That isn’t a job a citation chart can do for you.

Back at the SaaS startup, the trial disappears on Tuesday. Your tracker still shows a citation for best CRM with free trial. A buyer asks ChatGPT to compare options and gets last quarter’s offer with your client’s name attached. The client forwards you the screenshot and asks what the retainer is paying for. You own the visibility chart, but not the sentence that caused the complaint.

That is why I read every AEO pitch through the PPC scar. Much of what is sold as the best AEO tool for agencies is still a monitor. One 2026 agency report places the category on a monitor-versus-execute grid: monitoring is Profound’s strength, while execution is its weakness because showing patterns doesn’t publish a fix. Another comparison uses HubSpot’s AEO product as an execution example, pairing prompt monitoring with prioritized content fixes, rather than stopping at a report your team must act on. An agency teardown makes the commercial version of the same point: many tools were built for in-house brand teams, not agencies, with per-domain pricing, capped workspaces, and reporting that needs stitching together before a client sees it.

If you sell monitoring as though it includes control of the website, you’ve bought the Tuesday problem for every client on the roster.

Coda: the login matters more than the chart

If I were signing an AEO retainer to resell tomorrow, I’d start with access, not the tracker: CMS permissions, publishing rights in writing, technical fixes included rather than billed as change orders, and a log of each change and its reason. I don’t want to discover after a screenshot arrives that the person responsible for the answer has to file a request to edit the page.

That’s also how a reporting-only tool becomes a margin problem. Someone has to investigate the old offer, find every place it appears, persuade the client to approve a fix, get that fix made, and explain what happened. More dashboards can mean more unpaid work, the trap behind these seven AEO software myths that eat agency margins. I’ve done enough late-night account work to know the difference between seeing a problem and being able to solve it.

The search retainer timeline from rank reports to executed fixes describes the shift I’d rather sell: work changes on the site, the fix gets recrawled, and the agency can show what happened. On paper, I want ad-account and website context in one place, content and technical work handled alongside citation work, and actions logged clearly enough that I don’t have to reconstruct the night in a client report. Execution needs a path to the page. Otherwise the report only tells me what I was unable to prevent.

That is the shape groas builds its agency program around: connecting the client’s search work, carrying out changes, and keeping a record of the actions under the agency relationship. It answers the question I’d ask before reselling any AEO service. When the offer changes on Tuesday, who can put the right words on the page?

On the desk, the notebook is still open. $68 is crossed out. $140 sits underneath it. In the margin: refreshed design, with a question mark pressed hard enough to tear the paper.