How to find a good SEO company

You are buying something you cannot inspect. That is the whole difficulty. When I bought produce for restaurants, I could open the case. When you buy search work, you get a monthly PDF and a promise, and the feedback loop runs in months rather than days.

So the job is not to find someone who sounds impressive. It is to find someone whose work you will be able to verify while it is happening.

Where should I actually look for candidates?

Start with businesses like yours that are already winning, and ask them who does their work.

Not competitors, obviously. Ask the same business type in a city you do not serve. A dentist in Fresno will tell a dentist in San Jose exactly who they use and whether it was worth it, because you are no threat to each other. That single call is worth more than a week of searching.

Second source: your web developer, your bookkeeper, your POS rep. People who sit inside a lot of small businesses see which vendors survive and which ones churn.

Third source, with a caveat. People will tell you to hire the agency that ranks first for "SEO company near me." That is a weak signal at best. Ranking for your own service term proves someone spent effort on their own site, which is not the same as being good at yours, and plenty of strong operators are too busy servicing clients to compete for the most contested term in their industry. Use it as one data point, not a filter.

How do I tell competence from a good pitch?

Ask them to explain one thing you can independently check.

Pick something real about your own listing. Ask why you show in the map for one search and not another. A competent answer references your primary category, your proximity to the searcher, and your review activity. A weak answer references authority and optimization without naming anything you could verify by opening Google Maps.

If the mechanics of that ranking box are new to you, the map pack works on rules that are worth understanding before anyone pitches you, because you cannot evaluate a plan for something you have not seen the shape of.

Three questions with known answers, which makes them scoreable rather than impressive.

"Should I ask customers to mention the service in their review?" The right answer is no. Sterling Sky ran a controlled test on a Christmas tree farm with no other SEO activity, in summer, where six people left keyword-containing reviews against a base of 18. Rankings did not improve, and in the first instance they got worse. Google's own policy separately bans requesting that specific content be included in a review. A candidate who says yes is repeating a tactic that has been tested and failed.

"Will running Google Ads help my organic map rankings?" The right answer is no. Participation in Google Ads or other paid Google products ranks third on the low-scoring end of Whitespark's 2026 factor survey, among the things 47 local search practitioners believe do essentially nothing for local ranking.

"What does AEO mean and what would you do differently for it?" Google's own guidance says that from Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO, and it tells site owners to prioritize effective SEO strategies over AEO and GEO hacks. Google also states plainly that no third-party tool has access to its internal ranking or AI systems. A candidate who describes AEO as a separate product with its own dashboard is describing marketing. Most of what gets sold as AEO is the same work under a newer name, and how they answer tells you whether they follow the documentation or the deck.

The strongest question is still what they would decline. Ask what kind of client they turn down, and why. An operator with real capacity limits answers instantly. Someone who says they can help anyone is telling you they have never had to choose.

There is a longer set of these worth working through, and the questions that separate a good fit from a bad one are mostly not about SEO.

How do I score the plan they propose?

Here is the arithmetic. It takes fifteen minutes and it works on any proposal.

Whitespark publishes a Local Search Ranking Factors report in which 47 local search practitioners score 187 factors. The 2026 edition ranks the top local pack factors with combined scores: Primary Google Business Profile Category at 227, Proximity of Address to the Point of Search at 225, Keywords in the Business Title at 223, Physical Address in City of Search at 213, and Business Is Open at Time of Search at 189. Recency of Reviews sits at 164. Completeness of the profile at 147.

The bottom of the same list: Presence of Geo-tagged Photos at 23, Follower Count on Social Profiles at 25, Participation in Google Ads at 26, Keywords in Meta Descriptions Across the Entire Website at 33, Keywords in the Profile Description at 41.

Now take the candidate's proposed first ninety days and mark each item against that scale.

Ten proposed tasks. Count how many sit above 140 and how many sit below 50. If six of ten are below 50 and none of them touches the top five, you are looking at a template rather than a plan for your business.

The spread is what makes this useful. 227 against 23 is close to a tenfold difference in assessed impact between the first item and the last, which is much wider than the disagreement you would expect among 47 practitioners. The ordering survives the uncertainty even though the individual scores do not.

Grade the survey honestly when you use it. Darren Shaw, who runs it, says himself that the participants do not have access to Google's algorithm and that much of their opinion rests on observation, where correlation is not causation. He is right, and there is a documented case where the survey was wrong and a controlled test corrected it. Separately, a Search Engine Land analysis of 1.8 million Google Business Profiles across 4,209 categories found the exact-match category outranking adjacent categories for almost every category in the dataset, which is large-sample evidence for the item at the top of the list.

And if their proposal contains a table of click-through rates by local pack position, ask for the source. Google exposes no local pack position dimension in its performance reports or in Search Console, so no such table can be produced from platform data, and three of the four figure sets in circulation trace to studies that do not exist.

Does the agency need to be near me?

Usually no, and sometimes yes, and the yes cases are real and specific.

Most of this work is done in a browser. Location makes no difference to a category fix, a page rewrite, or a review cadence. Insisting on a local vendor narrows your pool for no gain in those areas.

But there are jobs where being able to walk the room is the whole job. Photographing your actual space rather than sourcing stock. Training your front counter staff to ask for a review in person, in your words, in front of your customers. Standing in your dining room at 7pm on a Friday to see why the layout confuses people. Sitting in your office for a quarterly planning session because your team will not engage with a video call.

There is a ranking argument for it too. Business Is Open at Time of Search entered Whitespark's 2026 top five, and getting your operating rhythm right, including seasonal shifts, holiday hours and the real time your kitchen stops taking orders, is something a person who has been in the building at 9pm gets right and a person reading a spreadsheet gets wrong.

If those are the things you need, hire someone who can drive to you. That is not sentiment, it is a correct reading of the work. An agency two hours away that shows up beats a better technician who never sees your operation.

What size of agency should I be looking for?

Match the shape to how you actually work, not to how big you want the invoice to look. Each option genuinely wins at something, and the honest version looks like this.

If you need same-day turnaround, coverage across time zones, or a name your board recognizes, go bigger. That is a legitimate requirement, not a weakness. If you want the operator on the phone and you are willing to wait a day for a non-urgent request, go small. If you need forty pages written this quarter and nothing else, the answer may be a production team plus a few hours of local review, which is a different purchase from an agency.

How the fee is structured tells you more than the size does, and the four common pricing models each shift risk somewhere different.

What to do this week

Build a list of three. More than that and you will not do the work on any of them.

For each one, do these four things. Ask for two clients in your industry you can call. Ask what they would decline. Ask how the work gets measured and what the number was when the client started. Then read one of their case studies properly, because case studies are written to be skimmed and they hide their weak points in the same three places.

Score each proposed plan against the factor list. Fifteen minutes, and it converts a persuasive document into a count.

If someone will not give you references, stop there. That is not a scheduling problem.

Questions about a shortlist you are already working through?

Email me at eric@seod.com with the names of the agencies you are considering, and I will tell you what to ask each one based on what they publish. I will not add myself to your list, and no sales followup comes back. If one of them is clearly the right fit, I will say so.

There is more on hiring and working with agencies if you want to keep reading first.