The AI SEO Agency Selection Guide — How to Tell Real AI Search Expertise from AI-Washed Traditional SEO

There are 1,900 monthly US searches for "ai seo agency" and roughly two hundred firms calling themselves one. Most are not. The phrase has become a marketing layer over traditional SEO retainers — a coat of ChatGPT-flavored paint on the same content-and-links playbook running since 2018. A small minority — possibly fewer than two dozen US firms — actually do the work of engineering a brand into the answer set that ChatGPT, Perplexity, Gemini, Microsoft Copilot, Claude, and Google AI Overviews quote.

This page is the honest selection framework. It names the five categories of agency you're actually choosing between, the ten red flags of AI-washing, the ten capabilities to demand, the five verification questions that separate real operators from pretenders, and the working list of agencies and platforms worth knowing — Rule27 included as one option, not the only one.

Rule27 is headquartered in Phoenix, Arizona. We are a registered AZ business with US-and-Canada client base, not a national agency with a UPS-box "Phoenix office." The AI SEO work itself is national — the generative engines don't care where your headquarters is, they care whether your entity graph, schema, and content treatment hold up to the citation logic of five different platforms. But the operator-being-real-and-reachable does change the experience of the engagement, particularly for AZ-based brands that want both local SEO depth and AI-search citation work in the same retainer.

For everyone else: the AI SEO work is delivered remotely, the audit and citation reports are written deliverables, and the monthly call runs on whatever video platform your team already uses. Geographic relevance does occasionally matter — Phoenix Business Journal, AZBigMedia, and ASU faculty pages are real citation sources we can place AZ-based clients into when the story fits — but the citation engineering work itself is national.

There are 1,900 monthly US searches for "ai seo agency" and there are roughly two hundred firms on the public internet calling themselves one. Most of them are not. The phrase has become a marketing layer applied on top of traditional SEO retainers — a coat of ChatGPT-flavored paint over the same content-and-links playbook that's been running since 2018. A small minority of firms — possibly fewer than two dozen across the US — are actually doing the work of engineering a brand into the answer set that ChatGPT, Perplexity, Gemini, Microsoft Copilot, Claude, and Google AI Overviews quote when a buyer asks a category-defining question.

This page is the honest selection framework. It names the five categories of agency you're actually choosing between, the ten red flags that mark an AI-washed shop, the ten capabilities a legitimate AI-search operator should demonstrate, the five verification questions that separate the two, and the working list of agencies and platforms in this market — Rule27 included as one option, not the only one. The goal is that you can walk out of any sales conversation with the vocabulary and the test questions to know whether the person across the table actually does this work.

The two products being sold under "AI SEO agency"

The sales decks look identical. The contracts read similarly. The pricing converges on roughly the same monthly range. But the two products being sold under the label "AI SEO agency" solve completely different problems, and most buyers don't find out which one they hired until the seventh monthly invoice.

The first product is an agency that uses AI tools to do traditional SEO work faster. The workflow runs through Surfer SEO or Clearscope for brief structuring, ChatGPT or Jasper for first-draft generation, MarketMuse or Frase for semantic gap analysis, and a junior editor at the end of the pipeline doing light cleanup before publication. The deliverables look like traditional SEO deliverables — keyword research, on-page optimization, content production, link building, monthly reporting on rankings and organic traffic. At the cheap end of this market the output is what the industry has settled on calling AI slop: shallow, recycled, footnote-free content that ranks briefly and gets demoted at the next helpful-content adjustment. At the careful end it produces respectable work, but it is still a traditional SEO engagement.

The second product is fundamentally different. It is the work of engineering your brand into the answer set the generative engines quote — ChatGPT, Perplexity, Claude, Gemini, Microsoft Copilot, and Google AI Overviews. The mechanics are not the same as ranking blue links. The measurement is not rankings or organic traffic; it is share-of-voice on a defined prompt portfolio. The deliverables include things most traditional SEO retainers don't touch: Wikidata claim editing, Knowledge Panel optimization, multi-surface citation monitoring, schema deployment specifically tuned for AI extraction, original-research publication, entity-graph audits, and monthly reporting that breaks share of voice out separately by surface.

The two products coexist legitimately. Rule27 sells both. But they need to be billed separately, scoped separately, and reported separately — and the central failure of the "AI SEO agency" market right now is that the cheaper of the two is sold under language borrowed from the more expensive one. That is the buyer's problem this page is designed to fix.

The five categories — what you're actually buying when you hire an "AI SEO agency"

The "AI SEO agency" market sorts into five categories. Each is a legitimate business in its own right; the harm comes from buyers thinking they're hiring one and actually hiring another.

Category 1 — Tools pasted as a service. A meaningful share of agencies billing themselves as AI SEO firms are essentially reselling Surfer SEO, Clearscope, MarketMuse, Frase, or Jasper with a thin services wrapper. The tools are excellent at what they do — content briefs, semantic gap analysis, on-page optimization scoring — but they are software products. They do not engineer your entity graph, they do not monitor citation share across the five generative surfaces, and they do not report AI-referred pipeline. The tell is usually a deliverable list that maps cleanly onto the software's feature list: "we'll deliver 8 SEO-optimized articles per month using our proprietary AI content optimization platform." The proprietary platform is Surfer.

Category 2 — AI-content-writing shops. The next category is agencies whose core service is producing content with AI assistance — sometimes carefully edited, sometimes barely touched. Pricing tends to be lower because the marginal cost of an AI-drafted article is close to zero, and the pitch usually emphasizes throughput ("30 articles per month for $5,000"). At the careful end, this work is fine; an experienced editor on top of a strong AI draft produces respectable mid-funnel content. At the careless end, this is the AI slop pipeline — Jasper drafts, offshore reviewers, Pexels stock photos, published at scale, ranks for a quarter, demoted at the next Google update. The structural problem for AI search specifically is that AI-generated content is not citation-worthy inside the generative engines themselves — the engines preferentially cite content with original analysis, primary research, named author credentials, and trackable entity signals, none of which a language-model draft has by default.

Category 3 — White-label resellers. A non-trivial slice of the "AI SEO agency" SERP is firms reselling another shop's work with a markup. The end-buyer doesn't know who actually does the citation tracking, who writes the content, who runs the Wikidata edits. The agency they signed with is a sales layer on top of a subcontractor pyramid. The tell is a refusal to name the team doing the work, vague answers about "our process," and project managers who can't speak to operational specifics.

Category 4 — Enterprise platforms with a services arm. Conductor, BrightEdge, Profound, and seoClarity are software platforms first; several of them have services arms that combine platform access with consulting. These are credible options for enterprise buyers with $200K+/year software budgets and the patience for 12-month implementations. They are not the right fit for a $5M to $100M revenue brand that needs a retainer it can evaluate monthly, but for genuinely large engagements they are real operators.

Category 5 — True AI-search operators. The smallest category. These are agencies whose primary work product is engineering a brand's presence in the AI answer set: defined prompt portfolios, multi-surface citation monitoring, entity-graph engineering, citation-density content treatment, original-research publication, monthly citation reports with the actual numbers. The category is small because the work is hard, the measurement is unfamiliar, and the talent market for someone who can credibly do all of it is thin. Rule27 belongs in this category. So does Onely. Searchbloom, eSEOspace, and a handful of others have made credible moves in this direction in the past 12 months.

Knowing which category you're actually shopping is the first step. Most buyers think they're shopping Category 5 and accidentally buy Category 1 or 2 because the marketing language is identical.

The 10 red flags of AI-washed agencies

If you encounter any three of the following in a single sales conversation, the agency is almost certainly Category 1, 2, or 3 wearing Category 5 vocabulary.

Red flag 1 — Guarantees AI visibility. No agency can guarantee AI citation any more than they can guarantee Google rankings. The engines change weekly. The citation logic is opaque. Any guarantee here is either a misunderstanding or a misrepresentation. The legitimate answer is some version of "we measure it weekly and we expect first measurable lifts in 60 to 90 days on long-tail prompts, with head-term movement at 180 days."

Red flag 2 — Talks "top spot in AI Overviews" and "keywords" instead of "citations" and "prompts." This is the highest-signal vocabulary tell. Real AI search work is measured in citations against a defined prompt portfolio, not in ranks against keywords. If the agency you're considering keeps reaching for the SEO vocabulary — "we'll rank you in AIO," "our keyword targeting for ChatGPT" — they're a traditional SEO shop with AI-washing on top. The vocabulary mistake is the workflow mistake.

Red flag 3 — Can't describe their measurement methodology. When you ask "how exactly do you measure my citation share," the legitimate operator can describe the prompt portfolio composition, the surfaces being polled (ChatGPT, Perplexity, Gemini, AIO, Copilot — each separately), the polling cadence, the citation-logging format, and the dashboard you'll log into. The AI-washed agency will say "we just track where you appear" or "we use proprietary tools." If they can't explain the measurement layer in detail, they don't have one.

Red flag 4 — Generic "AI SEO" playbook regardless of industry. AI search behaves differently across industries. B2B SaaS prompts skew technical and conversational; B2C commerce prompts skew transactional with Reddit weighting heavily; legal and medical prompts skew toward authority and credentialing signals. A real AI search operator will tailor the prompt portfolio and the surface emphasis to your category. A pretender will deliver a generic 8-article-per-month content plan regardless of whether you sell to dentists or developers.

Red flag 5 — No entity-graph work. Most brands look smaller to AI engines than they actually are because their Wikidata claims are incomplete, their Knowledge Panel is unclaimed, and their schema is unoptimized. The work of fixing this — Wikidata claim editing, Knowledge Panel optimization, schema deployment tuned for AI extraction — is central to real AI SEO and absent from AI-washed retainers. If the proposal doesn't name Wikidata, Knowledge Panel, or schema specifically, the agency isn't doing this work.

Red flag 6 — Uses AI to write content under your name. The most damaging AI-washed retainer pattern is the agency that runs ChatGPT or Jasper drafts under the client's byline, posts them, and reports content production volume as a success metric. The engines specifically demote this pattern. Worse, AI-generated content is structurally not citation-worthy inside the generative engines — they preferentially quote content with original analysis, primary research, and named human authors. AI-assisted research and brief structuring are fine; AI-replaced writing is the slop pipeline.

Red flag 7 — One "AI visibility" number instead of surface-by-surface scorecards. Every generative surface has different citation logic. ChatGPT browse mode behaves like a search ranker with freshness bias; Perplexity rewards recent citation-cited content; Gemini leans hard on the Google entity graph; Claude weights training-corpus authority; AIO mirrors organic top-10 signals; Copilot inherits Bing's Reddit weighting. A real AI search operator reports your citation share separately for each. An AI-washed agency rolls them into one number because separate reporting would expose how shallow their work is.

Red flag 8 — No prompt portfolio defined. Real AI search work begins with a prompt portfolio — typically 100 to 500 buyer-style questions specific to your category — that defines what "good" looks like for your retainer. If the agency you're talking to can't describe how they build a prompt portfolio, can't show you a sample, and can't articulate how the portfolio refreshes over time, they are not doing AI search work. They are doing content production with an AI sticker.

Red flag 9 — Tooling reseller. If their entire deliverable list collapses into "access to our proprietary AI search platform" plus light consulting, they are almost certainly reselling Profound, Conductor, AthenaHQ, or a similar enterprise tool with a markup. There is a place for that — Conductor's services arm is a legitimate operator — but the buyer should know they're buying software access plus consulting, not an operator-led retainer.

Red flag 10 — Hides pricing, hides the team, hides the methodology. Every "AI SEO agency" page on the first SERP that hides pricing, hides the team, and refuses to publish methodology is signaling that scope creep is the business model. The honest operators — there are several — publish prices, name their team, and publish their methodology because they want to be evaluated on what they said they'd do, monthly.

The 10 capabilities a real AI SEO agency must demonstrate

Reverse the red flags and you get the affirmative checklist. A legitimate AI SEO operator will show you the following on the first or second call, without making you ask twice.

Capability 1 — A defined prompt portfolio. 100 to 500 buyer-style prompts specific to your category, with a structured taxonomy (definitional, comparison, evaluation, commercial-intent, post-purchase) and a documented refresh cadence. The portfolio is the foundation; without it there is nothing to measure against.

Capability 2 — Multi-surface monitoring, separately reported. ChatGPT, Perplexity, Gemini, Google AI Overviews, and Microsoft Copilot, polled on a fixed cadence, with citation share broken out per surface. The reporting should also clarify which surfaces use browse mode versus training corpus, because the citation logic differs.

Capability 3 — Entity engineering. Wikidata claim editing, Knowledge Panel optimization, schema deployment (Organization, Person, Service, FAQ, HowTo, Article, Speakable, Breadcrumb), strategic brand-mention placement in the publications and communities that feed the training corpus.

Capability 4 — A citation dashboard you log into. Not screenshots in PDFs. Not "we'll send you the report monthly." A dashboard with prompt-portfolio share of voice over time, surface-by-surface scorecard, citation logs preserved for audit, and the ability for you to query the underlying data.

Capability 5 — Named human writers. Every page that goes out under your name is written and edited by a named human on the agency's team, with publicly verifiable credentials. AI assistance for research and brief structuring is fine; AI-replaced writing is the slop pipeline and the engines penalize it.

Capability 6 — Citation logs preserved for audit. When the agency claims a citation lift, you should be able to verify it. The raw citation logs — date, surface, prompt, response, brand mentioned, position — should be preserved and queryable, not summarized into a dashboard chart with no underlying detail.

Capability 7 — Surface-by-surface scorecard in monthly reporting. The monthly report breaks share of voice out separately by surface, identifies surface-specific wins and losses, and ties recommendations to surface-specific dynamics. "Up 12% across AI search" is not a real report. "Up 18% on Perplexity, flat on AIO, down 6% on ChatGPT browse — here's the surface-specific work this month" is.

Capability 8 — Published methodology. The agency publishes how they measure, how they build a prompt portfolio, how they prioritize entity work, and how they construct the monthly report. Methodology you can read before you sign is a higher trust signal than any sales claim.

Capability 9 — Original research capability. The single highest-citation asset class is original research — proprietary data the engines have no prior source for. A legitimate AI SEO operator can build and publish original research as part of the retainer. If they can only produce derivative content, they cannot move share of voice on head terms.

Capability 10 — Honest 60/90/180/365-day timelines. First measurable citation lifts at 60-90 days on long-tail prompts. Head-term citation share movement at 180 days. Sustained share of voice and measurable AI-referred pipeline contribution at 365 days. Any timeline shorter than this is either misrepresenting how the engines work or operating on a one-month measurement window that won't survive a quarterly review.

Five verification questions to ask any agency claiming AI SEO expertise

If you ask the following five questions in a sales conversation and the agency can answer all five with specific detail rather than category-marketing language, they are probably real. If they hedge, deflect, or reach for boilerplate on any of them, they're probably AI-washed.

Question 1 — "Walk me through how you'd measure my AI citation share. What's the prompt portfolio look like, what surfaces do you poll, and what dashboard do I log into?" The legitimate answer has concrete numbers ("100 to 500 prompts, refreshed quarterly"), names the five surfaces individually, describes the polling cadence ("weekly"), and offers a specific dashboard you can see before you sign. The AI-washed answer is vague — "we just track citations across the major AI platforms."

Question 2 — "Which Wikidata claims will you edit on my entity in the first 90 days, and how will you handle the Knowledge Panel?" If they don't know what a Wikidata claim is, the conversation is over. The legitimate answer describes the audit they'd run, the typical gaps for a brand your size, the approval process for Wikidata edits, and the relationship between Wikidata and the Knowledge Panel.

Question 3 — "Who specifically writes the content? Show me their bylines." A legitimate agency names the writers, points to their bylines on previous client work or on the agency's own publication, and confirms that no offshore content mill or pure AI generation enters the pipeline. The AI-washed answer is some version of "our content team" with no names.

Question 4 — "Show me a current client's citation dashboard with the prompts redacted." This is the single highest-leverage question because the legitimate operator has dashboards to show. They will redact the prompts and the client identity, but they can demonstrate the measurement layer working on real data. The AI-washed agency will refuse — usually with some version of "we can't share client data" — because there is no dashboard.

Question 5 — "What's your published timeline to first measurable citation lift, and what happens if we don't see one by then?" The legitimate answer is 60-90 days on long-tail prompts, with a documented escalation if the timeline slips. The AI-washed answer either promises too fast (30 days) or refuses to commit to any specific timeline.

Five questions, five minutes, and a clear read on whether the conversation is worth a second meeting.

The agencies and platforms worth knowing — an honest competitive landscape

The following are real operators or platforms in the AI search market as of mid-2026. The intent of this section is to give a buyer enough vocabulary and context to evaluate any of them against the framework above. Rule27 is included as one option, not the only one.

Profound is the purpose-built AI visibility monitoring platform and currently the enterprise leader in pure measurement. The platform tracks ChatGPT, Google AI Overviews, Perplexity, Copilot, and Gemini with real-time monitoring; it carries SOC 2 Type II compliance, has dedicated AI Search Strategist support, and offers the deepest analytics in the category. Profound is a tool, not an agency — the right buyer is an enterprise team that already has internal operators and wants the best measurement layer money can buy.

Conductor is the enterprise SEO platform that has invested heavily in AEO and now markets a direct head-to-head against Profound. The Conductor AI Insight Engine maps semantic gaps in AI conversations, ties citation work to pipeline revenue, and includes Content and Technical Agents that automate parts of the AEO workflow. Conductor has a services arm, real enterprise references, and the budget profile of a software-plus-consulting engagement — typically $200K+/year all-in.

Yotpo AI SEO is the commerce-specific play. Yotpo Reviews syndicates user-generated content into AI systems and feeds the engines the brand and product signals they look for in commerce queries. The right buyer is an established Shopify or commerce brand with existing UGC volume that wants to translate that UGC into AI citations specifically.

Mimvi is the New York-focused AI SEO agency, delivering AI-powered SEO and GEO services to NYC and Long Island businesses. The geographic specialism is real and the agency is credibly positioned for NYC buyers; the trade-off is the narrower service footprint relative to national operators.

Onely is a Polish-and-US technical SEO operator that has extended credibly into AI search. Their public content — including a frequently-cited "how to evaluate AI SEO agencies" piece — is among the most honest in the market, and their technical SEO bench is genuinely deep. The fit is best for technology and SaaS brands with complex technical SEO baselines.

Searchbloom publishes one of the more transparent AEO and GEO methodologies in the market. They sit in the "true AI search operator" category — small, focused, transparent — and are a credible alternative to Rule27 for buyers in non-Phoenix US markets.

Thrive Agency is a large full-service marketing firm that lists AI SEO as one service line among many. The Thrive content (notably the "10 Best AI SEO Agencies in 2026" listicle in which they rank themselves #1) is heavy marketing relative to operational specificity. They are a credible Category 2/4 hybrid for buyers who want a single agency for paid, SEO, and AI search under one contract.

eSEOspace markets aggressive citation-lift claims ("75-85% average increase in AI citations within 90 days") and pioneered a "unified approach" framing. The marketing is hot enough that verification is the buyer's responsibility — ask for the dashboards behind the claim.

Coalition Technologies and Level Agency are large full-service shops with AI SEO service lines. Both are credible at scale; both fit the "enterprise platform-style services" category more than the lean operator category.

SEO.ai and Ranked.ai are tool-as-service plays — software products that perform AI SEO work largely autonomously, with light human review. They are credible for small businesses that want the cheapest possible AI SEO retainer and accept the trade-off of less human judgment in the loop.

Passionfruit, Embarque, NoGood, Skale, Single Grain, and Siege Media appear repeatedly in third-party "best AI SEO agencies for startups" lists and represent the content-led tier of the market. Each has real operators on staff; each fits a different size and stage of client.

Rule27 is the Phoenix-based AI SEO operator with transparent retainer pricing on the website ($3,500 audit, $4,500/month Foundation, $8,500/month Growth, $15,000+/month Enterprise), named operators on every engagement, a citation dashboard every client logs into directly, and month-to-month commitments after a 30-day satisfaction window. We are not the only honest operator in this market. We are one of them.

How Rule27 fits in this market

The Rule27 differentiation, said plainly: we publish what most agencies hide.

Transparent pricing. Retainer tiers and audit pricing are on the website. The Audit is $3,500 one-time, the Foundation Retainer is $4,500/month, the Growth Retainer is $8,500/month, and Enterprise starts at $15,000/month. Every retainer is month-to-month after a 30-day satisfaction window. No 12-month contracts. We have nothing to hide behind.

Named operators. You'll know the human who runs your prompt-portfolio monitoring, the editor who reviews your re-engineered content, and the person who walks you through the monthly citation report. We do not hide the team behind an account-management layer.

A citation dashboard you log into. Not a PDF. Prompt-portfolio share of voice over time, surface-by-surface scorecard across ChatGPT, Perplexity, Gemini, AIO, and Copilot, citation logs preserved for audit, and the underlying data queryable by your team.

Named human writers. Every page that goes out under your name is written and edited by a named human on the Rule27 team. We use AI tools for research and brief structuring; we do not use AI to draft content under client bylines. The engines penalize that pattern, and we don't ship work the engines penalize.

We separate "agency that uses AI" from "agency that gets you cited by AI." Most of the SERP conflates the two. We explicitly bill them separately, because they are different products solving different problems.

We are AZ-based, headquartered in Phoenix. Our clients sit across the United States and Canada. We are not the only honest operator in this market — we'd point a NYC buyer to Mimvi, an enterprise buyer to Conductor or Profound, a content-led startup to Passionfruit or Embarque, and a Shopify commerce brand to Yotpo before pretending Rule27 is the answer for everyone.

If the framework above describes the agency you'd want to hire, the shortest path is the free AI Search Visibility Audit at the top of this page. We'll measure where your brand currently appears across ChatGPT, Perplexity, Gemini, and Google AI Overviews against your three closest competitors. 48-hour turnaround, real document, no auto-bot output. Even if you don't hire us, you walk away with a measurement baseline that didn't exist before.