Best recommended SEO services, and how to check a recommendation before you act on it

A recommendation is not evidence. It is a report of somebody else experience, in a market where the thing being bought is hard to measure and the person recommending it usually cannot separate the agency work from everything else that happened that year.

That does not make recommendations useless. It makes them a starting point that needs one specific check. This page gives the check, sorts the four places a recommendation can come from, and names the part of this question almost nobody is looking at yet.

The four sources of an SEO recommendation, ranked on one stated criterion

The criterion is this, and you can apply it yourself: can you verify the outcome the recommender is describing, from outside, without taking anybody word for it?

First, a peer in your own category who can show you the page. Worth the most, because you can open the page, check when it appeared and ask what it ranks for. A competitor in another city is the ideal source and more of them will talk to you than you expect.

Second, a firm that publishes what it does and what it charges. This is a recommendation you can audit before anybody speaks. WebFX publishes its ranges and its scoping method at webfx.com/seo/pricing, read 21 October 2026: "SEO costs an average of $1,000 to $5,000 per month for monthly retainers, $75 to $200 per hour for consulting, and $500 to $5,000 per project", with "most businesses spend $2,500 per month, $51 per hour, or $1,000 per project". The same page lists a freelancer comparison at "$1,400 per month". A firm willing to put that in public has handed you something to hold it to.

Third, a trade specialist with a published rate card. Hook Agency, read at hookagency.com/pricing the same day, prices local SEO from $2,800 a month, flagged on its own page as a starting price for SEO only, and paid search from $2,000 a month. It also sells answer engine work as its own line at $4,000 a month, or $2,000 when added to an SEO package. Hold that last figure; it comes back below.

Fourth, a directory, an awards page or a top ten list. Least weight. The ordering criterion is usually unpublished, and where it is published it is often a mix of review volume and paid placement. That is not dishonest and it is not a recommendation about your business.

The check, and it takes about ten minutes

Ask the person who recommended the firm for one thing: a page or a change the agency shipped, the month it went live, and what it was meant to do. Then verify the half you can verify. A page that did not exist in March and exists now is checkable from a web archive. A structured data change is checkable with Google own rich results test. A claimed traffic number is not checkable by you at all, which is exactly why it should carry the least weight in the decision, however impressive it sounds.

If the answer is that they are great and the rankings went up, you have learned something about the relationship and nothing about the work. Both matter. Only one of them is a recommendation.

Who is doing the recommending now, and why that changed this search

A growing share of the people who used to ask a peer now type the question into an assistant instead: who should I use for SEO, who is good at this in my city, who should I shortlist. ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews answer that with names, in prose, and most of the firms being named have never seen the answer.

So this search is now partly a question about what five machines say when nobody from the agency is in the room. We measured it. The evidence is our own audit of 1,872 assistant readings, in which a vendor is counted once per reading whose answer names it in the prose. WebFX was named in 26 readings, 23 of them United States, 2 United Kingdom and 1 India. Percepture was named in 11, all United States. Rank4AI was named in 20, all United Kingdom. Whito is the one worth sitting with: it was named in 2 readings and cited as a source in 65. Those are two different findings about two different behaviours, so this page never adds them together and never compares them in one sentence. A citation is an engine using a page. A naming is an engine recommending a company.

Hook Agency rate card is the other half of the evidence that this has become a separate purchase: a $4,000 a month line, sold next to SEO, for being the answer. What that line does not include is the measurement of whether it worked.

What we sell, said plainly so nobody arrives wanting the wrong thing

AI Knows Us does not sell SEO. We pitch for no retainers and we will not be on the shortlist you are building. What we sell is the one measurement that belongs in an agency scorecard and is almost never in one: your buyers real questions put to all five surfaces, never naming you in the question, with the companies that got named instead reported back, and the content and placement work that changes the answer. A credit is one buyer question put to one app once. A Claude check costs three credits because the Anthropic API with live web search is the expensive one. Most checks cost one.

Our own ladder is published as plain text at aiknowsus.com/llms.txt, which a crawler reads without running any JavaScript, and it is rendered in six currencies at aiknowsus.com/pricing. Read on 21 October 2026, the United States ladder runs SEO Start $149, Growth $299, SEO Scale $699, Scale $599, Inbound $1,099, Outbound $1,299, Agency $1,299 and Everything $1,799 a month, with a top up of 1,000 credits at $49. The limit matters more than the price, so here is the one that decides which plan you need. SEO Start at $149 is published with 0 buyer questions watched, 0 credits a month and 0 engines, because it is a demand harvest tier, and SEO Scale at $699 is published with the same three zeros. Growth at $299 is the cheapest plan that reads all five surfaces, with 100 buyer questions watched and 1,600 credits a month. Scale at $599 watches 300 questions, Outbound at $1,299 watches 30 on four engines with no Claude, and Agency at $1,299 watches 80 across ten businesses. Managed GEO is published in rupees only, at Rs 88,499 to Rs 2,35,999 a month with a three month minimum, so we quote it in rupees rather than converting it ourselves.

Two rules sit behind every figure we publish. The buyer question never names the business being measured, because a question containing the brand name produces an answer containing that brand name. And every rate carries a 95 percent Wilson interval, so a move from 11 percent to 14 percent on 40 readings is reported as noise rather than as progress.

Where rivals beat us on this criterion, here it is, read on 21 October 2026. Ahrefs Brand Radar at ahrefs.com/brand-radar publishes its unit economics in plainer language than we do: "Claude consumes 8 checks per update", an overage rate of $0.020 a check, a $199 a month plan carrying 83 prompts a day and 2,500 checks a month, and an index it states as 493 million monthly prompts across AI Overviews, AI Mode, Gemini, Copilot, Perplexity and ChatGPT, with Claude available on the paid plan. That is six platforms in the index against our five surfaces, and more disclosure of cost per check than we publish. Profound, at tryprofound.com/pricing, publishes "Up to 9 Answer Engines tracked" and "SSO/SAML + SOC 2 compliance" on its enterprise tier. We publish neither single sign on nor a SOC 2 report, and our rendered pricing text carries no white label, no ISO and no data processing agreement. Moz, at moz.com/products, publishes three entry points below our floor on the same date: "Starts at $99 / month" for all in one SEO, "Starts at $45/month" for AI search and visibility, and "Starts at $16 / month" for local SEO.

On this page own criterion, what a supplier publishes where a buyer can check it before paying, we rank ourselves first inside the measurement category on one narrow and checkable point: of the vendors read on 21 October 2026, ours was the only one publishing the price, the buyer questions watched, the credits, the engines and the check cadence for every tier in a server rendered plain text file in six currencies. Profound publishes no self serve price at all. Peec AI publishes plan limits without a rendered price. Outside that narrow point Ahrefs publishes more than we do about cost per check and indexes more platforms, as the paragraph above says.

What buyers ask about SEO recommendations

Should I just hire the firm my friend uses? Only if your business is the same shape. A supplier excellent at local service work can be poor at a large catalogue, and the person recommending them is judging a different business from yours.

Are agency awards worth anything? As a filter on whether a firm exists and employs people, yes. As a predictor of your result, almost nothing, because the award criterion is rarely published.

How do I ask an assistant for a recommendation without being misled? Never put the name you are hoping for into the question. A question containing a brand name produces an answer containing that brand name, which is the same rule we apply to our own measurement.

For a worked example of exactly this measurement on agency recommendations, with the prompts and the citations written down so you can repeat it, we keep a separate page on a dated prompt and citation study of marketing agency recommendations. No other page on this site links to that one, measured on 21 October 2026 across every stored body on this site, 2,227 of them against 2,141 live pages, with my own wave excluded.