Red Flags in an SEO Proposal

Read for commitments, not promises

Every proposal contains two kinds of statement, and they usually sit in different sections. Promises are about results: what your traffic will do, what will happen to your rankings, what revenue will follow. Commitments are about work: what will be examined, what will be produced, by whom, by when, and how it will be measured.

The first category is worth nothing, and not because providers are dishonest. Nobody outside Google operates the ranking system. Google's own documentation states there is no assurance that changes to a site will produce noticeable impact, and that if more deserving content exists it will continue to rank. A promise about position is a promise about something the promiser does not control.

So read a proposal by crossing out every sentence about outcomes and seeing what remains. If most of the document disappears, you are holding a brochure. If what remains is a scope you could hold someone to, you are holding a proposal.

The patterns below recur. Some are disqualifying on sight; others are worth one clarifying question, and the answer decides it.

Guaranteed positions, traffic or revenue

The clearest one, and Google states it directly: no one can guarantee a number one ranking, and you should beware providers who claim to, who allege a special relationship with Google, or who advertise a priority submission. That guidance is published on Google's own developer documentation and is the strongest single citation in this entire subject.

Watch for the softened variants, which are more common than the blunt version. “Guaranteed first-page placement” without saying for which terms — usually terms nobody searches for. “Guaranteed traffic increase”, which can be satisfied by buying visits or by targeting terms so obscure that new traffic is worthless. “Guaranteed results or you do not pay”, where the results are defined afterward by the party defining them. A confidence-backed refund offer, where the refund turns out to be a service credit obliging you to stay.

The reason this matters beyond the specific claim: a provider offering a guarantee has told you their business model requires methods with predictable short-term effects. In search, the methods with predictable short-term effects are largely the ones that violate policy, and the consequence of that lands on your domain rather than theirs.

Secret, proprietary or unexplained methods

Google's wording is unambiguous: be careful if a company is secretive or will not clearly explain what it intends to do, and insist on understanding every change made to your site. That last clause is the operative one. If you cannot understand a change, you cannot approve it, and you certainly cannot undo it after the provider is gone.

The phrasings to notice: a proprietary algorithm or scoring system that will grade your site; a methodology described only by an acronym; a private network of publishing sites that cannot be named; an assurance that the technique is safe but the details are commercially sensitive.

Nothing in mainstream search work is a trade secret. The mechanics are documented publicly by Google, in web standards, and in openly published research. What differs between practitioners is judgment, prioritization and craft — none of which requires concealment. A genuine methodology can be explained to a technically literate buyer in ten minutes, and the good ones enjoy explaining it.

A related tell in the 2026 market: products sold as AI-specific markup or special files required for AI features. Google states plainly that no new machine-readable files or special structured data are needed to appear in these features, and that there are no additional technical requirements beyond being indexed and snippet-eligible. A proposal selling the opposite has a citation available against it.

Links that are bought, quotaed, or not shown to you

Buying a link is not automatically a violation. Google's spam policies are specific: paid links are acceptable when qualified with rel="nofollow" or rel="sponsored". The violation is the undisclosed paid link. So the question to a provider is not whether money changes hands but whether the resulting links are qualified, and whether you will see every one of them.

Flags in the document itself:

• A monthly link quota. A guaranteed number of links per month implies a supply, and a supply implies a purchase or a network. Genuine link acquisition is lumpy and unpredictable because it depends on other people deciding to link.
• Unnamed placement sites, or a promise to reveal them after signing.
• Anything described as private, exclusive or a network in relation to publishing.
• Guest posting at volume priced per unit, which is a purchase however it is described.
• Directory or search engine submission at scale. Google's guidance names submitting your site to thousands of search engines as a typically useless exercise.
• Routine monthly disavowing. Google's own guidance says most sites will not need the disavow tool. Selling it as a standing service line sells work against a problem you probably do not have.
• A required link back to the provider. Google's wording: you should never have to link to an SEO.

Deliverables written as activities rather than artifacts

This is the most common weakness in otherwise respectable proposals, and it is the one that costs buyers the most money because it does not look like a problem.

An activity can be satisfied by almost nothing. “Ongoing optimization.” “Monthly SEO maintenance.” “Content strategy.” “Technical improvements.” “Link building.” “Competitor monitoring.” None of these can be marked complete or incomplete by anyone, which means at renewal time there is no factual basis for the conversation — only impressions.

An artifact either exists or it does not. “A technical findings document with each item specified to implementation level, prioritized by affected traffic and effort, delivered by a stated date.” “A monthly change log listing every modification made, with dates.” “A rebuilt account structure with the campaign map documented.” “Twelve pages, each against a named query set, drafted and delivered for approval.”

The same applies to volume-based deliverables, which look concrete and are not. Ten articles a month is a manufacturing commitment, not an outcome, and it is possible to deliver ten unhelpful ones indefinitely. If a proposal is priced on volume, ask what determines the subject of each unit and who decides whether it was worth producing.

A simple test: for every line in the deliverables section, ask whether a neutral third party could look at the month's work and say yes or no. If not, that line is not a deliverable.

No baseline, and no definition of what will be measured

A proposal that does not begin by recording where things stand is a proposal that cannot be evaluated later, and the absence is sometimes deliberate.

What should be in there: a defined set of pages and queries; impressions, clicks and average position captured on a stated date before work starts; the conversion definition, agreed and identical across channels; and a statement of which metrics will be reported and which cannot be known.

That last item is a strong positive signal when present. An honest measurement section says out loud that Search Console retains 16 months and caps the interface at 1,000 rows; that rare queries are suppressed, so filtered totals never sum to unfiltered ones; that average position means the position of your topmost result for a query rather than an average across your pages; that rank tracking samples a synthetic query and is a diagnostic rather than an outcome; and that no source currently reports clicks from AI features, so any vendor claiming AI traffic attribution is modeling rather than measuring.

The opposite signal is a single composite score — an “SEO health score” or “visibility index” — presented as the headline metric. That number reports the vendor's own tool's opinion, moves when the tool changes, and cannot be reconciled with anything in your own systems.

Nobody named, and terms designed to be hard to leave

Two commercial patterns worth reading carefully, neither of which Google addresses, so both are argued on their own merits rather than by citation.

No named person. If the proposal describes a team's collective experience but names nobody who will touch your account, you cannot evaluate what you are buying. This is not an accusation — a pitch team differing from a delivery team is a normal agency structure. But notice that some firms advertise a minimum experience standard for their consultants rather than naming who you get, and others promise explicitly that you will speak to a senior practitioner. Those promises exist because buyers learned to ask. So ask: who, specifically, and will they be on the calls.

Terms that make leaving expensive. Twelve-month minimums with no review point. Automatic renewal with a notice window that has usually just closed by the time anyone remembers. Long notice periods. Early termination fees. Work product or documentation withheld on exit. Advertising accounts held under the provider's own management identifier with no transfer arrangement discussed, or analytics and Search Console properties you do not own.

A minimum useful term is legitimate — search work has a real lag and judging it in six weeks measures nothing. A twelve-month lock-in with no exit is protecting the provider's revenue forecast. The difference is whether there is a review point at which either side can stop without penalty.

What a proposal should look like instead

Since the list above is negative, here is the positive version. A proposal worth signing contains most of these and is unembarrassed about the ones it cannot include.

• A restatement of your problem that is better than the brief you supplied, showing that somebody thought about your situation specifically.
• A hypothesis and a method — what they think is wrong and how they intend to confirm or eliminate it, rather than a list of improvements applicable to any site.
• Deliverables as artifacts, each with a date, and an explicit exclusions section. The exclusions are the most useful part of any proposal.
• A named person, with their actual role and involvement stated.
• A baseline plan and a measurement section that states its own limits.
• Dependencies on you, listed with owners and timing — developer capacity, content approval, access, a decision-maker.
• Cost drivers explained, so you can see what would make the number larger or smaller.
• Exit terms and ownership, plainly stated.
• At least one thing they recommend you do not buy. Rare, and worth a great deal when present.

Nothing on that list is exotic, and everything on it is checkable before you sign — which is the whole point. The proposals that go wrong are almost never the ones a careful reader could not have seen coming.

Frequently Asked Questions

Published 2026-08-11