How to choose an SEO agency in 2026

Choosing an SEO supplier is difficult for a specific reason: the thing being sold is hard to evaluate before you buy it, and the industry has a long history of selling activity instead of results. A convincing pitch deck costs almost nothing to produce. A competent technical diagnosis does not.

This is a guide to telling the two apart before you sign.

The thing most businesses get wrong

Businesses usually evaluate agencies on presentation quality, client logos and price. All three are weak signals. Presentation quality measures the sales team. Client logos measure the sales team's historical success, not the delivery team's. Price measures almost nothing at all without knowing what it buys.

The strong signal is much simpler: can they tell you something true about your site that you did not already know? A supplier who has genuinely looked will have found something specific — an indexing problem, a cannibalisation issue, a competitor taking a query set you should own. A supplier who has not will talk in categories.

The questions that actually reveal competence

"What would you do in the first ninety days, and why in that order?" The single most useful question. A good answer is specific to your site and explains the sequencing: barriers first, because they cap everything else. A weak answer is a list of deliverables with no reasoning about order.

"What have you found on our site already?" Anyone serious will have spent an hour looking. If they have not, they are selling a process rather than solving your problem.

"Who will actually do the work, and what is their day job?" You want to know the ratio of senior to junior time, and whether the person in the room will be the person on the account. This is the most common gap between what is sold and what is delivered.

"Show me a campaign that did not work, and what you learnt." Every real practitioner has several. An inability to name one means either inexperience or a rehearsed answer.

"How will you report, and what will lead the report?" If rankings and traffic lead, the supplier is optimising for what is easy to show. Qualified leads and pipeline should lead; traffic explains them.

"What happens to the work if we leave?" Content, research and documentation should be yours outright. Watch for anything that lives only in the agency's tooling.

Warning signs

• A guaranteed number one ranking. Google does not sell positions. A guarantee means either ignorance or a deliberately meaningless target (ranking first for your own brand name, for example).
• Refusal to explain the method. "Proprietary" is sometimes real, but a supplier who cannot explain what they will change on your site is a supplier you cannot supervise.
• Links as the whole strategy. Buying links from networks is detectable and the penalty lands on you, not on them.
• Reporting that never changes. If the monthly report is structurally identical every month, nobody is making decisions from it.
• No interest in your sales process. Someone optimising for revenue needs to know what a good lead looks like. Someone optimising for traffic does not.

Agency, freelancer or in-house?

In practice many businesses end up with a hybrid: an in-house owner who sets priorities and holds context, with external specialists for technical work and content production. That works well, provided the internal owner has enough authority to get development changes shipped.

How to run the selection

1. Audit first. Know your own numbers before anyone pitches. It changes the questions you ask and makes it obvious who has actually looked.
2. Brief identically. Give every candidate the same information and the same question. Differences in the answers are then meaningful.
3. Ask for the ninety-day plan in writing. Not a proposal — a plan.
4. Take a reference from a client who left. The most informative conversation available, and rarely requested.
5. Start small if you can. A scoped technical project before a twelve- month retainer lets both sides find out how the other works.

When to change supplier

Change when the reporting stops producing decisions, when you cannot get a straight answer about what was done last month, or when the work has quietly become maintenance. Do not change because of a single bad quarter — particularly not in the first six months, which is when the investment is front-loaded and the return has not arrived yet.

If you do move, get the handover in writing: access to every property, documentation of what was changed and why, and the research files. A supplier who makes that difficult has told you something about the relationship.