Lourdes Paul Agilan · 2026-08-26

Choosing an AI SEO Agency: A Buyer's Framework

Note on verification: agency positioning quoted below was read from each company's own website on 26 August 2026 and is quoted verbatim. Named agencies are category examples, not recommendations and not a ranking, for reasons set out in the section on listicles. Pricing is separated into three tiers by evidence quality: prices an agency publishes itself, minimums self-reported to directories, and survey estimates. We did not source any figure from a "best AI SEO agencies" page. One clarification we should make up front: we found no FTC enforcement action against a GEO, AEO or AI SEO vendor, and the FTC cases referenced below are structurally analogous, not precedents in this category.

Read this before your first call: "AI SEO agency" is currently one label on at least three different products. There are traditional SEO shops that added an AI service page, standalone GEO and AEO specialists, and full programs combining measurement with execution. A fourth group uses "AI SEO" to mean they use AI tools to produce SEO work faster, which is a claim about their workflow, not an outcome for you. The term doesn't tell you which one you're being pitched. Working that out is your first job on the call.

This is a framework, not a ranked list. If you want the reasoning behind that choice, skip to the section on listicles.

The term is unsettled, and the buyers are behind the practitioners

Three separate surveys asked what people call this work, and got three different leading answers.

Fractl surveyed 342 search practitioners in November 2025 and found 42% personally used "GEO", with AEO and LLMO trailing (Fractl, Nov 12, 2025). SEOFOMO's survey of more than 200 senior SEOs in September 2025 found the most common answer was different again: 36% said clients simply call it "AI search optimization", 27% stuck with SEO, 18% said GEO (Search Engine Land).

Then Fractl ran the same question at 343 marketing decision-makers rather than practitioners, and the picture inverted: 81% still call it SEO internally, only 27% have formally adopted any term beyond SEO, and 42% have actively decided against doing so. When those buyers go looking for help, 70% would search either "AI search optimization" (46%) or "SEO" (24%) (Search Engine Land, Aug 24, 2026). Same research firm, near-identical sample size, opposite answer, because it's a different population. These aren't a time series and shouldn't be read as one. They're a supply-side and demand-side gap: the practitioners renamed the work, and the buyers didn't.

That gap is the commercial context for everything below. When vocabulary is this unsettled, a label is free to acquire and expensive to verify.

The four things "AI SEO agency" can mean

Type 1: an SEO agency with an AI service page. The tell is the deliverable list. Victorious' answer engine optimization page frames AEO honestly as "best viewed as an extension of search engine optimization (SEO), not a replacement", then lists deliverables that are a standard SEO retainer with "AEO" prefixed to three of the lines (Victorious). Amsive sells "SEO + AEO" with AEO as one line item among technical, ecommerce and local SEO (Amsive). Neither is dishonest, and both are explicit that AEO extends SEO rather than replacing it, which is more than many competitors manage.

The industry's own candour about how common this is comes from Terry Williams of First Rank, quoted in Search Engine Land: "We're sprinkling GEO-related tasks into SEO campaigns" (Search Engine Land, Jan 21, 2026). The structural evidence agrees: SE Ranking's agency survey found 61% of agencies planned to offer AI-engine optimization, and 43% intended to integrate it into existing services rather than build anything new (SE Ranking). On Clutch, most firms listed under generative engine optimization self-report GEO at around 20% of their service mix (Clutch). GEO is a fifth of the typical listed firm's work, not a practice.

Type 2: born-in-AI specialists. Firms whose entire offer is AI search. AEO Engine calls itself "the AI SEO agency that makes you the answer in AI Overviews, ChatGPT, and Google" and, unusually, publishes its prices (AEO Engine). iPullRank has gone further and renamed the discipline entirely, selling "relevance engineering", defined as "the art and science of improving visibility for any search surface" across AI Overviews, TikTok search, ChatGPT, Perplexity, Amazon and app stores (iPullRank).

Type 3: measurement plus execution programs. The diagnostic here is that the deliverables are measurement artifacts, not tactics. Seer Interactive's GEO page lists a Prompt Development and Tracking Plan, an AI-Driven Traffic Conversion Analysis, AI Analytics Benchmarks, and a Traffic Predictive Modeler (Seer Interactive). Whatever else you think of that offer, you can tell what you'd receive each month.

Type 4: "we use AI to do SEO." Single Grain's AI SEO page leads with "AI SEO that plans, writes & ranks - 90+ hours/month saved" (Single Grain). The headline metric is hours saved by the producer, not visibility gained by the client. That's a legitimate service; it just isn't the one most people searching "AI SEO agency" think they're buying.

The confusion isn't hypothetical, and it lives inside individual agencies. WebFX runs two distinct services under AI branding: one defining "AI SEO services" as "the use of AI to improve a website's visibility in organic search results" (WebFX), and a separate generative engine optimization service for AI search visibility (WebFX). Both are reachable from the same navigation. EWR Digital manages both meanings on a single page, describing AI SEO as content automation in one paragraph and "being surfaced, cited, and trusted by Large Language Models" in another (EWR Digital). If a buyer lands on the wrong page, they buy a workflow tool believing they bought a visibility outcome.

How to tell the difference on a sales call

Most of the general evaluation criteria we'd apply to any agency, process transparency, named staffing, real case studies with baselines, sane contract terms, are set out in How to Choose a GEO Agency for B2B SaaS, and we won't restate them. These six questions are the ones specific to sorting the four types apart.

"Which of these are you: a tool, a service, or both, and what arrives in my inbox each month?" A report, or finished work. This one question separates types 1 through 4 faster than anything else.
"Show me the exact prompt list you'd track for us. Who wrote it, you or your tool's default?" Aleyda Solis' guidance is that vendor default prompt sets "won't automatically reflect your products, audiences, markets, competitors, constraints or business priorities", and that a real starting point is 30 to 50 commercially relevant prompts organised by product line, audience, market and journey stage (Aleyda Solis, Jun 8, 2026).
"How many times a day does each prompt actually run?" Not how many prompts. How many runs. Research on AI answer variance puts the required sample somewhere between 7 and 100 runs per prompt depending on whose methodology you accept, and almost no vendor publishes the number. An agency that has never thought about this is reporting noise with a decimal point.
"Do you report ChatGPT, Gemini, Perplexity and AI Overviews separately, or blended into one score?" Blended is a red flag. Kevin Indig found 91% of citations appear on only one engine (Growth Memo, Jul 27, 2026), so a single blended number averages across near-disjoint populations.
"Which of your deliverables would Google's own documentation call unnecessary?" Google's guidance states there are "no additional requirements to appear in AI Overviews or AI Mode", that "you don't need to create new machine readable files, AI text files, or markup", and that "there's no requirement to break your content into tiny pieces for AI to better understand it" (Google Search Central). If llms.txt files are on the proposal, ask what evidence supports them. Ahrefs found 97% of published llms.txt files were never fetched across 137,210 domains (Ahrefs, Jun 15, 2026).
"What's the leading indicator you'd move before share of voice?" Seer's answer is brand-fact accuracy, on the argument that visibility is worth little if the AI describes you wrong. Any coherent answer is fine. No answer means the program starts at the dashboard.

And the sentence to listen for, from Jeremy Moser of uSERP: "If a GEO service does not openly tell you that success in AI visibility is 80 percent good fundamental SEO, they are selling you snake oil" (Digiday, Mar 2026). Wil Reynolds of Seer puts the same test differently: "Anybody who tells you they've got a silver bullet for GEO, I'd run the other way."

Red flags specific to this category

Guaranteed AI rankings or citations. Structurally impossible, and both platform owners say so in their own documentation. Google: "indexing and serving isn't guaranteed" (Google Search Central). OpenAI: "placement is not guaranteed" (OpenAI Help Center). Note the distinction that makes this useful rather than absolutist: a guarantee scoped to traffic measured in your own GA4 and Search Console, against a disclosed prior baseline, is a real commercial commitment somebody can actually keep. A guarantee of citations or placement is not. Ask which one is on the table, in writing.

Quantified performance claims with no methodology. Several agencies in this category advertise figures like "920% average AI traffic growth" with no sample size, no baseline definition and no published method. The FTC's standard, in place since 1984, is that advertisers must "have a reasonable basis for objective claims before they are disseminated" (FTC), and its Operation AI Comply cases against firms making unsubstantiated quantified claims came with the line "there is no AI exemption from the laws on the books" (FTC, Sep 25, 2024). To be precise, and we'd rather be precise than alarming: those cases targeted other industries and we found no enforcement action against a GEO vendor. The substantiation standard is general; the precedent isn't specific.

Paid brand mentions sold as authority building. Search Engine Land documented the emerging practice in detail: private blog network mentions sold at 10 to 15x typical backlink cost, placement fees around $250 per mention, brands placed on topically irrelevant domains, mass Reddit posting with aged accounts, all framed to clients as "partnership building" (Search Engine Land, Jun 26, 2026). Lily Ray's comment in the same piece is the historical read: "going back to Penguin in 2012, paid mentions on low-quality sites have been treated as spammy link building."

Programmatic comparison pages at scale. Ray again, on the tactic of programmatically generating a thousand near-identical "us versus every possible competitor" pages that are "all 80% similar": that is "scaled content abuse" under Google's existing spam policies, and negative competitor claims in those pages are "asking for a lawsuit for sure" (PPC Land, May 14, 2026).

Tactics that testing says don't work. The peer-reviewed C-SEO Bench, presented at NeurIPS 2025, evaluated the named "GEO techniques" circulating in this market and concluded that "most current C-SEO methods are not only largely ineffective but also frequently have a negative impact on document ranking", with conventional SEO outperforming dedicated techniques (Puerto et al., 2025). That is the strongest single piece of evidence a buyer can bring to a pitch.

A single unqualified upside claim. Digiday reported case studies showing AI visibility gains alongside a 66% organic traffic loss. If a vendor's pitch has no downside case at all, they haven't run into one yet, or they aren't telling you.

What it should cost

There is no credible published survey of GEO or AI SEO retainer prices specifically. Anyone quoting you a confident market rate is estimating. Here is what can actually be verified, in descending order of evidence quality.

Prices agencies publish themselves. AEO Engine publishes a ladder from $1,597/month (Growth) to $10,000/month (top Dominate tier) (AEO Engine). WebFX publishes GEO service tiers starting at $3,000/month and running to $68,250/month at enterprise (WebFX). Almost nobody else does: we found no published pricing for Siege Media, Victorious, iPullRank, Seer Interactive, NoGood, Terakeet, Amsive, Ignite Visibility, Skale or Relixir.

Directory minimums. Clutch's generative engine optimization category, as of August 2026, shows minimum project sizes from $1,000 to $50,000 across listed firms, with hourly rates from $50 to $300+ (Clutch). A 50x spread inside one directory category is the finding. There is no midpoint worth quoting.

Survey estimates, with heavy caveats. SE Ranking's agency survey reports an average of $937/month for finalised AI-engine optimization services. Two caveats that make that number nearly unusable on its own: 78% of that sample were agencies with fewer than 10 people, and the authors themselves note it may not reach statistical significance. For general SEO, Ahrefs' survey of 439 providers found an average monthly retainer of $2,917 (Ahrefs), though that data is now roughly two years old.

The honest summary: a small-agency AI search add-on sits near $1,000/month, a specialist program sits in the low thousands, and enterprise engagements run to five figures, and the ranges overlap so heavily that price tells you almost nothing about what you'll receive. Which is exactly why the deliverable question comes first.

Why this isn't a "best AI SEO agencies" list

Because the format doesn't work, and we have the data.

Our own research into self-promotional listicles found the pattern repeatedly: these pages reliably earn citations and reliably fail to get the publisher recommended, and past a certain volume they promote whoever already dominates the category. Ahrefs ran a controlled test across 34 self-promotional pages and 9,886 AI answers and found the AI used the article as a source while naming a competitor instead, in one case recommending a competitor's event in 43% of answers (Search Engine Journal, Aug 18, 2026, sponsored by Ahrefs).

There's a quality signal in these pages too. One prominent "10 Best AI SEO Agencies" page ranks its own publisher first, was authored by that publisher's Head of Growth, discloses no methodology or conflict of interest, and leaks an unrendered [current_month_year] template placeholder into its own metadata so the page auto-stamps a fresh date. Siege Media, Single Grain, First Page Sage, Directive, Go Fish Digital and NoGood all publish "best GEO agency" listicles that include themselves. These are lead generation assets. Read them as such, and don't take facts from them.

FAQ

How do I know if a GEO agency is legitimate or just a rebranded SEO shop? Ask what changed. A rebrand adds a service page; a real capability adds a measurement system. Concretely: does the proposal contain a named, disclosed prompt set built for your business? Does it report engines separately? Does it state a sampling rate? Does it name a leading indicator that isn't share of voice? Can they tell you which of their own deliverables Google's documentation says is unnecessary? A rebranded shop will answer the first question with tactics and the rest with silence. One caveat that keeps this fair: being a repositioned SEO agency is not itself disqualifying. Lily Ray, whose "GEO grifters" critique is the sharpest in the industry, works at an agency that sells AEO as an extension of SEO, which is precisely her stated position. The problem isn't the pivot. It's the pivot that adds nothing under the label.

What questions should I ask before hiring? The six in the sales-call section above, plus the general agency diligence in our buyer's framework. If you only ask two: what exactly arrives each month, and show me the prompt list. Those two answers sort almost every vendor in this market into the right box.

Is AI SEO actually different from SEO, or is it the same work renamed? Both, and the two camps are arguing about different things. Google's own documentation says there are no additional requirements or special optimizations for AI features, and the C-SEO Bench results say most dedicated GEO tactics underperform ordinary SEO. That's the tactics argument, and it's largely correct. Mike King of iPullRank argues the opposite from a different premise: that the skillset (retrieval theory, RAG pipeline analysis), the measurement, and crucially the budget treatment are genuinely different, and that calling it "still SEO" means "more scope, same budget, no new authority" (iPullRank, May 18, 2026). Both parties have a structural incentive: Google benefits from discouraging an optimization industry aimed at its AI surfaces, and specialists benefit from the work sounding new. The defensible position is that the tactics overlap heavily while the measurement, prompt research and multi-platform coverage genuinely don't, and that distinction is exactly what separates a bolt-on from a program.