What SEO Consulting Costs

Why there is no number on this page

A published price for search work is either a lie or a description of a product. If it is a product — a fixed list of deliverables shipped every month regardless of what your site needs — then the number is real, and so is the reason it can be quoted before anyone has looked at your site. If it is genuinely diagnostic work, the figure depends entirely on things nobody knows before looking, and any number stated in advance is a guess wearing a suit.

So instead of a figure, this page gives you the structure. Every driver below moves the cost in a direction you can predict, and by the end you should be able to tell whether a quote you are holding is high, low, or simply describing different work from the one next to it. That is more useful than a range, because ranges published by providers are chosen to make the provider look reasonable.

It also avoids a familiar waste of everyone's time. Buyers who arrive at a number first tend to shop for that number, and what they find is the version of the work that fits it.

What the available pricing data actually is, and how old

Worth knowing before you accept anyone's characterization of the market, including mine. The best primary dataset in public is a survey of 439 search service providers, published in December 2023 and updated in August 2024. That is the dataset everyone quotes, directly or indirectly, and it is not 2026 data. Treat it as structure and as a floor, not as current pricing.

Two findings from it are worth carrying even without the figures attached. The first is that the distribution is enormously wide, and the modal band sits well below the average, which means the average is being pulled up by a minority of large engagements. Anyone quoting the average as typical is describing a market that most providers are not in.

The second is the one that surprises people. Consultants did not come out cheaper than agencies. On average, independent consultants charged slightly more per month than agencies did, and freelancers sat in a different band entirely, far below both. The folk model — agency expensive, consultant cheap, freelancer cheapest — is contradicted by the only decent dataset available. What differs between a consultant and an agency is cost structure, not fee: no account manager layer, no sales layer, no junior execution layer. You are buying fewer people, not cheaper hours. The survey is published in full if you want to see the distributions yourself.

The drivers, roughly in order of how much they move the figure

Scope drives cost. Company size barely does, and any provider pricing off your revenue rather than your site is pricing off what they think you can pay.

• Whether implementation is included. The single biggest swing, and the one most quotes are ambiguous about. Diagnosis and specification is one price; someone making the changes is another entirely.
• Whether content production is included. The second biggest. Content is manufacturing, priced per unit, and it does not belong in the same line item as consulting.
• Site size and template count. Not page count alone — the number of distinct templates, because that is the actual unit of work. A hundred thousand URLs across nine templates is less work than four thousand URLs across forty.
• Competitive difficulty of the queries you want. Not a mystical quality. It is observable: who currently holds those positions, how established they are, and whether the result page still sends clicks at all.
• Number of markets and languages. Each locale multiplies the validation work, and international annotation errors are found by checking every pair.
• Platform. Some platforms make some work impossible rather than expensive. Hosted commerce platforms that cannot expose server logs or custom sitemaps set a ceiling on what any provider can do, which is a cost fact even though it looks like a technical one.
• Whether a migration or a penalty is in play. Both are bounded, both are urgent, both carry more risk than routine work.
• Reporting and stakeholder load. Unglamorous and real. Four stakeholders in three time zones is a genuine cost, and honest providers price it rather than absorbing it and resenting it.

Diagnosis and execution are different purchases

The most common reason two quotes differ by a wide margin is that they are not for the same thing, and neither document says so plainly.

Diagnosis is finding out what is wrong and specifying the fix precisely enough that a competent developer could build it. It is bounded, it produces a document, and it ends. Execution is somebody doing the work: writing the code, publishing the pages, running the account. It is unbounded by nature, because there is always more of it.

A quote that includes execution will be larger and should be. A quote that excludes it is not cheaper — it has moved a cost onto your team, and if your team has no capacity, that cost lands anyway, later, as an unactioned document. When comparing two proposals, the first thing to establish is which side of that line each one sits on, and the second is who is expected to do the building.

What a low quote usually means

Most low quotes are not deceptive. They are describing a different product, and there are four ordinary explanations before you reach a dishonest one.

Junior execution. The work is real but performed by someone early in their career, with senior review that may or may not exist. Sometimes fine, particularly for well-specified implementation. Rarely fine for diagnosis, which is where experience does most of its work.

A templated deliverable. The same audit structure applied to every client, populated by a tool. Cheap to produce, occasionally useful, and identifiable by findings that do not name your URLs.

A fixed-hours package. A set number of hours a month regardless of what the site needs. The hours are real; the fit is accidental.

A loss leader. The audit is priced below cost because it exists to sell the retainer. You can spot this in the document: the findings that are cheap to fix are underweighted, and every path leads to ongoing work.

Then there is the version that should end the conversation. A price far below everything else, attached to promised rankings or link volume, is frequently funded by work that breaches published spam policies — undisclosed paid links, or pages generated at scale without adding value. Google is unambiguous that "no one can guarantee a #1 ranking on Google," and warns against providers promoting link popularity schemes or claiming a special relationship. If a low price arrives with a guarantee attached, the guarantee is the tell, not the price. Google's guidance on hiring an SEO is worth reading before any proposal meeting.

What a high quote should come with

Expensive is not the same as good, and the check is symmetrical. A larger figure should be accompanied by specifics that a smaller one cannot offer: named work rather than categories, a stated sequence, an explicit statement of who performs each part, and a definition of what would count as failure.

Two questions separate a well-priced senior engagement from an expensive one. Who actually does the work, by name, and what happens if they leave? And what portion of this figure is diagnosis, what portion is implementation, and what portion is account management? Providers who cannot break that down are not necessarily hiding anything, but they have not thought about it, and you will be paying for that too.

How to compare two proposals on like terms

Rewrite both onto the same six lines before comparing any figures. It takes twenty minutes and it settles most decisions.

1. What is diagnosed, and against what evidence — crawl data, server logs, Search Console, analytics, or a tool's default report.
2. What is implemented, by whom, and what happens if your developers are unavailable.
3. What is produced — the artifact you keep, not the outcome you hope for.
4. What is excluded. Content production, development time, tooling and advertising budget are the usual omissions, and any of them can exceed the fee itself.
5. How it will be judged, stated before the work starts, using measures neither side gets to choose retrospectively.
6. What happens at the end — what you own, what transfers, and whether the arrangement continues by default or by decision.

Two proposals that look four times apart on price frequently turn out to differ on lines two and four rather than on quality. Occasionally the expensive one turns out to be the cheaper purchase, because the other one quietly assumed a development team you do not have.

What is worth paying for, and what is not

Worth paying for: diagnosis by someone with the experience to know which findings matter, specification precise enough to be built, judgment about sequence, and verification that what shipped behaves as intended. All four are hard to fake, and all four are what actually changes outcomes.

Not worth paying for: reporting that repackages numbers you already have access to, tool subscriptions billed as strategy, monthly deliverables that exist to justify the invoice, and anything sold on a promised position. Also not worth paying for — repeating diagnosis you have already bought. If a recent competent audit exists, say so, and expect it to come off the proposal.

The last thing to know is that cost is not the risk here. The risk is paying a fair price for work aimed at the wrong problem, which is invisible for months and costs the same as the right work. That is why the diagnosis is the part to buy carefully, even though it is the smaller number.

Frequently Asked Questions

Published 2026-08-11