More clients now arrive with a plan an AI assistant wrote and ask us to carry it out. We do not refuse and we do not rubber-stamp. This post sets out the keep, change, drop review we run on each line of such a plan, with a worksheet you can use yourself and an illustrative plan worked through.
Why the plans read so well
A prospect used to arrive with a problem and a vague sense of the fix. Increasingly they arrive with a specific, confident plan: a channel mix, a content calendar, sometimes exact tactics. That is a real improvement, because the first conversation can start from specifics rather than education.
It also brings a new failure. The plan is fluent, and fluency reads as authority. Nothing in the tone of the answer tells you which lines fit your business and which would fit any business in the category.
Three reasons a fluent plan can be wrong for you
It answers the general question. “How should an online store approach SEO?” has a good average answer. Your store’s biggest problem may not be on that list at all, because the assistant could not see your rankings, your site or your competitors.
It can be out of date. Search and ad platforms change faster than training data, so a line can describe how something worked a year or two ago. The answer gives no signal that it is stale.
It does not know your constraints: budget, existing contracts, the platform you are on, brand rules, who on your side can approve and ship work. None of that was in the prompt, so none of it shaped the plan.
The keep, change, drop review
We take the plan line by line and put each line through three checks. You can run the same review with access to your analytics and Search Console.
- Fit. Does your own data show this is a problem on your site? A speed recommendation needs your field data to show slow pages; a content recommendation needs a gap in the searches you want to win.
- Currency. Does the platform still work this way? Check the line against the platform’s own current documentation, not a summary of it.
- Constraints. Can you do it with your budget, stack, contracts and approvers? A sound tactic you cannot ship is not a plan.
- Verdict. Passes all three: keep. Right idea, wrong form or order: change. Fails fit or currency: drop.
- The missing line. Finally, ask what your data shows that the plan never mentions. That line is often worth more than everything on the list.
- When
- Passes fit, currency and constraints
- What it usually means
- The assistant got it right, and we say so
- When
- Right idea, wrong form or order
- What it usually means
- A generic idea applied without your constraints
- When
- Fails fit or currency
- What it usually means
- It answers the general question, or it is stale
An illustrative plan, reviewed
Illustrative: a plan of the kind an assistant writes for a mid-sized online store, with the verdicts our review would typically reach. The verdicts depend on the site’s own data, which is the point.
| Line in the plan | What the review checks | Typical verdict |
|---|---|---|
| Improve page speed | Field data for the templates that earn revenue, not a lab score for the home page | Keep if those templates fail; otherwise lower its priority |
| Add FAQ schema to every page for rich results | Google limited FAQ rich results to well-known government and health sites in 2023 | Drop as a rich-results tactic; keep FAQs only where they help readers |
| Target the highest-volume keyword in the category | Who ranks for it now, and whether its searchers buy | Change: target the searches one step closer to a purchase |
| Publish several blog posts a week | Who writes and approves them, and whether the topics map to what buyers ask | Change: fewer, deeper pieces against a topic map |
| Build backlinks through guest posting | Whether the links would be earned or bought, and the risk under search spam policies | Change: earn coverage with something worth citing |
| (Not in the plan) | Two pages competing for the same search; a category the store sells but has no page for | Add, often first |
Where the change is large, the plan needs rebuilding rather than editing. For content, that is our topic map and strategy sprint, which takes 2 to 3 weeks on our content marketing rate card. Where most lines survive, we recommend keeping the plan and building the work around it.
When the AI plan is right, and when you do not need us
Often most of the plan is reasonable. The failure is rarely a bad idea; it is a generic idea applied without your constraints. Where a line passes all three checks, we keep it and tell you plainly that the assistant got it right.
And sometimes the review is all you need. If the plan passes and you have people to carry it out, run it yourself. A focused SEO short focused review can confirm the riskiest lines without a programme. If the plan’s first recommendation is to “build an agent”, what that word actually means decides the brief.
If your worry is the reverse, that an assistant recommends a competitor when buyers ask about you, the AI visibility audit describes how to capture that evidence before changing anything.
Written by
Parvessh Agarwal
Founder and Chief Executive, Vinove
Parvessh founded Vinove in 2004 with one idea: build focused technology companies that earn their place in real markets. His focus is on the long game, on companies that survive and serve real customers long after launch.
LinkedIn →
