How the work runs
Why the order never changes
The sequence below is the same for every site. That's deliberate, and it isn't laziness.
Reinventing the process for each client sounds attentive. In practice it hands the vendor the power to decide what gets looked at, and the things they would rather not look at quietly don't get looked at. A fixed order means the awkward findings surface whether or not they suit me — including the ones where the honest answer is that the site is fine and your problem is somewhere else entirely.
It also means you can check the work. You know what stage three involves before you pay for stage one, and you are in a position to tell me I skipped something.
The Diagnostic, stage by stage
Crawl and render
I crawl the site the way a bot does, then render a sample of pages the way a browser does — because those two often disagree, and the gap between them is usually where the damage is. Status codes, redirect chains, canonicals, indexability, what the HTML actually contains before JavaScript runs, and how the thing behaves on a mid-range phone on an ordinary connection rather than on a laptop with fast internet.
OutA list of technical faults, each ranked by how many URLs it touches.What your queries actually return
Then I take the searches your buyers really type — their phrasing, not the keyword-tool version — and look at what the results page gives them today. Whether an AI answer sits above everything else. Who it quotes. Whether that answer resolves the question completely, which changes what a click is even worth for that query, and sometimes changes whether the query is worth chasing at all.
OutA query-by-query picture of who currently owns the answer, and where there is room.Content and authority, read honestly
Then the pages themselves. Whether anything on the site is genuinely quotable — specific, sourced, structured, attributable to someone real — or whether it is the same paragraph every competitor has already published. Alongside that, the link profile: what points at you, whether any of it is worth anything, and whether some of it is going to become a problem later.
OutA short list of pages worth strengthening, and the ones better merged or deleted.The report, and an hour on the phone
Everything above, numbered, in priority order, with a cost to fix beside each item in money or in developer hours. Then a call to go through it. I would rather you argue with the priorities than nod along — you know your business, your budget and your constraints, and I have only been looking at it from the outside.
OutA document your developer can work from. It's yours whether or not we continue.
What I won't promise
Most of what makes this trade look bad is promised at the sales stage. So here is the list, before you pay for anything.
A ranking. Ever.
I don't guarantee a ranking, because nobody selling SEO controls the index. Anyone promising you position one is either guessing, or quietly planning to rank you for a phrase nobody searches so the promise technically lands.
A date by which you will be at a position
Same problem with a calendar attached. The moment a deadline is promised, the work bends toward whatever shows movement fastest, which is rarely whatever is actually worth doing.
That an AI will cite you
I can make a page easier to quote: clear claims, real specifics, clean structure, an author who exists. I cannot make a model choose it. These systems are retrained and rebuilt without notice, and no vendor has a lever on the output — including the ones advertising that they do.
Links I bought, exchanged or built on a network
No paid link placements, no private blog networks, no reciprocal schemes at scale. They breach the search engines' own spam policies, and when it goes wrong the penalty lands on your domain rather than on my invoice. That is your asset being risked for my convenience.
A contract you cannot leave
Month to month, no minimum term, no notice period engineered to make leaving expensive. Everything I produce for you is yours when you go, in a format that opens without me in the room.
That a tool score is a business result
A green audit score, a ninety-something on a speed test, a visibility index ticking upward — none of those are money. I'll show them where they're a useful signal, but I won't put one at the top of a report as though it were the outcome. The outcome is enquiries, or it isn't anything.
A report that quietly omits the bad month
Some months go backwards. You'll get why I think it happened and what I would change, including the times when the honest answer is that I don't know yet and I am still working it out.
How this actually behaves over time
There are three different clocks running, and treating them as one is how people end up disappointed.
Technical fixes can move quickly. Unblock a template that was stopping pages rendering, or unwind a redirect chain, and the change can register as soon as those pages get recrawled — which might be days, or might not be, depending on how often anything crawls your site at all. This is the part where speed is genuinely possible. It is also the part most likely to be half-done already by the time I arrive.
Content and authority move slowly and unevenly. A page published today is competing with pages that have had years to accumulate links and reputation. Work in this direction usually looks like nothing is happening, and then a group of pages starts moving together. Months rather than weeks. Anyone quoting you weeks for this is selling you the technical work with a content label stuck on it.
Generative visibility is the least predictable of the three, and I'd be lying if I dressed that up. Whether a model quotes your page depends on what it was trained on, what it retrieves at the instant somebody asks, and product decisions made by companies who change them without telling anyone. What can be influenced is the input: being specific, being structured, being the source rather than the summary of someone else's source. What cannot be influenced is the decision. Any vendor drawing you a projection for this is drawing it from imagination.
One more thing, because it always comes up. A drop is not automatically something you did. Core updates, a competitor relaunching, seasonality, a results-page layout change that pushed every organic result further down the screen. Part of the job is telling the difference before anyone panics and rewrites a site that was working.
Start where everyone starts
The Diagnostic is the only way in. Fixed scope, a published price, and a report you keep either way.