How to Spot a GEO Agency Scam: 8 Red Flags in AI Search Visibility Pitches and a Pre-Contract Checklist
How to screen GEO agency pitches that promise guaranteed AI search visibility, claim control over algorithms, or flaunt big-brand logos. A table turning 8 red flags into documents to demand before signing, the evidence from Google's official guidance and SparkToro's variability study, a comparison of independent verification tools, and a contract clause checklist.
The quickest way to spot a GEO agency scam comes down to one word: "guaranteed." A pitch that guarantees visibility in AI search is technically impossible, and in 2026 Korean news outlets and industry newsletters issued one warning after another about exactly this kind of selling.[1][3] The trouble is that many marketers are commissioning GEO work for the first time and don't know what to ask. This guide lists eight warning signs of risky sales tactics and converts each one into a document you can demand before signing. The short version: what separates a sound agency from a risky one is not a polished proposal but whether it can hand over its raw measurement data.
What Is a GEO Agency Scam? A 30-Second Definition
- GEO (Generative Engine Optimization) is a marketing strategy for getting a brand and its content cited and recommended as a trusted source in generative AI answers.[1]
- A guaranteed-visibility GEO pitch is a sales approach that promises a specific position or a set number of citations in AI answers as the outcome and charges for it.
- A GEO red flag is a signal that an agency is steering you toward a contract with unverifiable promises or references you cannot confirm.
- A verifiable GEO proposal is one that documents the target queries, measurement period, denominator, formula, and deliverables before the contract is signed.[2]
Why No One Can Guarantee AI Search Visibility
Two public sources are enough to show why guarantee pitches are wrong. One is the platform operator's official position. The other is the result of repeated experiments.
Google's documentation on AI Overviews and AI Mode says plainly that indexing and serving are not guaranteed (Google Search Central, 2025).[5] The same page says there are no additional optimizations needed to appear in AI features, and no new machine-readable files or special schema.org structured data are required.[5] Google's advice on hiring an ordinary SEO says "no one can guarantee a #1 ranking on Google" and tells site owners to look elsewhere if a provider promises one (Google Search Central, accessed 2026).[6] If classic search rankings can't be guaranteed, a promise of placement in AI search, where every answer is generated fresh, doesn't hold together.
Experimental data points the same way. SparkToro analyzed results from 600 volunteers who ran 12 prompts through ChatGPT, Claude, and Google AI Overviews a combined 2,961 times (SparkToro, 2026).[7] Ask the same question 100 times, and the chance that ChatGPT or Google's AI returns the same brand list in two responses was under 1 in 100; the chance of matching the same order as well was about 1 in 1,000.[7] Under those conditions, a "top placement screenshot" may simply be the best of many attempts.
The academic numbers don't support guarantees either. The paper that coined the term GEO reported that optimization methods raised visibility by up to 40% on its own benchmark (arXiv, 2023).[8] That figure is a ceiling, and it was measured in an experimental setting. An agency that cites it to promise results for a specific brand is misreading the research.
8 GEO Agency Red Flags and the Documents to Demand Before Signing
The table below collects warning signs that came up repeatedly in Korean press coverage and industry columns.[1][2][3][4] The column that matters is the last one. When you spot a signal, don't argue; ask for these documents on the spot.
| Red flag | Example pitch | Why it's a problem | What to demand before signing |
|---|---|---|---|
| Guaranteed visibility | "Guaranteed top placement in ChatGPT," "50 AI citations guaranteed" | Google explicitly states it does not guarantee visibility in AI features | Target query list, sources currently cited per query, citation rate from repeated measurement with the formula |
| Claims of algorithm control | "We control the AI logic," "We have connections inside the portals" | AI answers change from run to run, and the platforms themselves do not guarantee visibility | The lead's published research, training materials, and work history |
| Big-company logo references | A row of big-company logos on the website | Press reports describe logos posted without clear evidence of any real engagement | Client's responsible department, project period, anonymized before-and-after measurement data |
| Prepayment and sign-now pressure | "Prices go up soon," "Only one slot left" | Takes away the time you need to verify | A quote with installment payments and interim reporting dates |
| One-screenshot results | A single screen capture, or a capture of a query containing the brand name | AI answers change every run, and branded queries surface the brand by default | Raw repeated-measurement results for non-branded queries, with the denominator |
| Technical promises that contradict official guidance | "Install llms.txt and you'll get cited," "Guaranteed AI citations through schema" | Google says AI features need no special file or special schema | Diagnosis and fix records for crawl access, indexing, and content issues |
| Substituting SEO metrics | Reports only keyword rankings and visitor counts | These are different metrics from citations in AI answers | Citation rate per AI service, GA4 AI assistant session baseline |
| Data and ownership lock-in | Hands over one-off screenshots only, nudges you to transfer domain ownership | Nothing is left after termination: no measurement history, no assets | Clause for exporting raw data, domains and accounts kept in the client's name |
Risky Sales Tactics Actually Reported in Korea
The reported Korean cases fall into two groups, and the contract structure adds a third risk.
The first is fake references. An employee at a financial company told Hankyung that they found their own financial group's logo displayed as a client on an AI marketing agency's website; when they asked about it, the agency blocked them instead of replying (Hankyung, 2026).[1] A newsletter covering the same case reported that after an industry insider made similar cases public, the agency removed some of the logos (Oz's Knowledge Talk, 2026).[3]
The second is flaunting authority and manufacturing urgency. According to the report, which cited the Q-Letter newsletter from Korean marketing community i-boss, some firms cited elite-university backgrounds as proof of AI expertise, and pushed for prepayment with "prices go up soon" and "only one slot left right now" (Hankyung, 2026).[1] In the same article, a PR agency executive warned against firms that promote results with a single screenshot or guarantee short-term wins from optimization that takes time.[1]
The third is not a reported case but a risk built into the contract structure: data lock-in. If you receive only screenshots without the raw measurements, if content goes up on a domain in the agency's name, or if ownership of your domain moves to the agency, the moment you terminate you lose both the measurement history and the domain where the content lives.
Practitioner columns add a few more. GEORank lists as warning signs SEO agencies that rebrand as GEO without changing anything, agencies that answer "if you do SEO well, GEO follows naturally," and mass publishing of content that just strings keywords together (GEORank, 2026).[4] The same column estimates that changes show up in about one to two weeks on Perplexity and three to six weeks on ChatGPT and Gemini.[4] That is one agency's working estimate, so treat it as a reference point for filtering out pitches that promise "results within a week," not as a hard benchmark.
Comparing Tools to Verify Agency Claims Independently
Screening for red flags gets you only so far; the final call should rest on numbers. Don't rely on the agency's report alone. Measure the same queries yourself, or cross-check them with a separate tool. Why tools produce different scores is covered in Why AI visibility tools disagree.
| Measurement tool | Type | Public facts | How to use it to check an agency |
|---|---|---|---|
| GA4 AI assistant channel | Free analytics tool | Classifies AI assistant referrals as a default channel | Baseline for AI referral sessions and key events before and after the contract |
| Search Console generative AI reports | Free analytics tool | Reports URL impressions within Google's AI features | Check the agency's claims about visibility in Google AI |
| Profound | Global monitoring | New York, USA | Cross-check citation rates for target queries |
| Peec AI | Global monitoring | Berlin, Germany | Cross-check share of mentions versus competitors |
| Otterly.ai | Global monitoring | Austria, offers a GEO Audit | Repeated checks on a small query set |
| Scrunch AI | Global monitoring | Salt Lake City, USA, AXP | Cross-check citation rates for target queries |
| Next-T OPTIGEO | Korean solution | Hanam, Gyeonggi; positions itself as a GEO automation platform | Check citations for Korean-language queries |
| Across GPTO | Korean solution | Positions itself as monitoring multiple LLMs | Check citations for Korean-language queries |
| BOIDA BVI | Korean solution | Presented as connecting measurement to diagnosis and improvement work (AVING, 2026) | Check citations for Korean-language queries, compare the basis for diagnoses |
No tool simply confirms an agency's report as-is. The point is to run the same query set in at least two places and see whether the direction matches. When BVI, the BOIDA product run by Designovel, was exhibited at MWC 2026, it was presented as a system that takes a URL or product name, analyzes visibility across major generative AI services, diagnoses where the brand falls short of competitors, and carries through to improvement work (AVING, 2026).[9] For how Korean service types differ, see the Korea GEO agency comparison and selection guide.
Pre-Contract GEO Agency Checklist (5 Steps)
In practice, work through these steps in order. If the documents for a step don't materialize, don't move on to the next.
- Screen for guarantee language: Search the proposal and any call recordings for "guarantee," "control," "exclusive," and "one slot left." If you find any, request the documents in the last column of the table above.[1][2]
- Get raw case data: Ask for anonymized before-and-after work logs and raw measurement data. Exclude screenshots of queries that contain the brand name.[2]
- Confirm the measurement design: Get the measurement period, query list, target AI services, denominator, and key event definitions in writing. Say 60 attempts produce 2 errors, leaving 58 valid responses, and your URL is cited 7 times; the citation rate is then about 12.1%. Someone at the agency should be able to show that calculation and explain it (digitalmarketer.co.kr, 2026).[2]
- Test technical audit skills: Ask for records of diagnosing and fixing AI crawler access, indexing, and content rendering problems. Check any pitch that leans only on llms.txt or schema against the complete llms.txt guide and Google's guidance.[5]
- Turn it into contract clauses: Agree up front on the deliverables list, interim reporting cadence, and terms if targets are missed (GEORank, 2026), and spell out that raw measurement data can be exported and that domains and accounts stay in the client's name.[4]
Whether the price is reasonable is a separate question from this checklist. Price ranges and what they include are covered in the GEO agency monthly cost guide and the GEO pricing and contract structure FAQ. A low price doesn't make an agency safe, and a high one doesn't make it legitimate. The test is whether it can show its diagnostic accuracy and measurement design on paper.
The One Sentence That Separates Sound Agencies from Risky Ones
A sound agency promises process and measurement, not results. A risky agency promises results and hides how it measures. You can surface the difference in the first meeting with one request: "Show me the target query list and the formula first." If you're still deciding between running GEO in-house and outsourcing it, reading GEO agency vs. in-house vs. solution alongside How to evaluate a GEO company will speed up the decision. New to GEO itself? Start with What is GEO, and find candidate firms in Recommended GEO companies.
Key Takeaways
- Guaranteed AI search visibility does not exist. Google does not guarantee visibility in its AI features (Google Search Central, 2025).
- AI answers change with every run, so a single screenshot is not evidence (SparkToro, 2026).
- Risky sales tactics reported in Korea include unauthorized logos, flaunted authority, and prepayment pressure (Hankyung, 2026). Data lock-in is a separate risk to close off in the contract.
- When you see a red flag, ask for the target query list, raw repeated-measurement data, and the formula.
- The contract should cover data and domain ownership, deliverables, reporting cadence, and termination terms.
Related companies
- 넥스트티 (Next-T, OPTIGEO)SEO, GEO, AEO 컨설팅, 자동화
- 디자이노블 (Designovel, BOIDA)AI 패션 테크, 생성형 AI, GEO
- 보이다 (BOIDA)생성형 검색 최적화(GEO) 솔루션, AI 가시성 측정
- 비즈스프링 (BizSpring, LOGGER)웹, 데이터 분석(MarTech), GEO 컨설팅
- 어크로스 (Across, GPTO)AEO, GEO 답변 최적화 엔진
- Otterly.aiAI 가시성 모니터링 툴
- Peec AIAI 가시성 모니터링 플랫폼
- ProfoundAI 가시성 모니터링 플랫폼
- Scrunch AIAI 가시성 모니터링 플랫폼
Frequently asked questions
- No. Google states that indexing and serving are not guaranteed for its AI features (Google Search Central, 2025), and the odds of an AI repeating the same brand list for the same question are under 1 in 100 (SparkToro, 2026). Instead of a guarantee, ask for citation rates per target query from repeated measurement, along with how they are calculated.
- Not on their own. AI answers change from run to run, and a screenshot of a query that includes the brand name will show the brand almost by definition. Ask for the raw results of target queries without the brand name, run several times, together with the denominator.
- Treat it as inconsistent with official guidance. Google says AI Overviews and AI Mode require no new machine-readable files and no special schema.org markup (Google Search Central, 2025). Schema work makes page content easier to read; it is not a way to guarantee citations.
- Skip the logos and ask for the client's responsible department, the project period, and anonymized before-and-after measurement data. The press has reported a case where a company asked whether a collaboration was real and the agency blocked it instead of answering (Hankyung, 2026).
- Yes. Set a traffic baseline with the GA4 AI assistant channel and the Search Console generative AI reports, then measure target-query citation rates separately with global tools such as Profound, Peec AI, and Otterly.ai or Korean solutions such as Next-T OPTIGEO, Across GPTO, and BOIDA BVI.
- It commits to process and measurement, not outcomes. Before the contract, it documents the target pages, the changes to be made, the schedule, how completion will be confirmed, and the measurement period, query list, and denominator, and it agrees up front on what happens if targets are missed.
Q.A GEO agency says it can guarantee top placement in ChatGPT. Should I believe it?
Q.Do the AI answer screenshots an agency shows me count as proof of results?
Q.How should I read a pitch that promises AI citations from installing llms.txt or doing schema work alone?
Q.How do I check a GEO firm that displays big-company logos as clients?
Q.Are there tools to verify an agency's claims myself before signing?
Q.How does a legitimate GEO agency commit to results?
Sources
- [1] ↑챗GPT 답변에 광고 노출? 사기성 마케팅 주의보 — 한국경제
- [2] ↑GEO 측정 기초: AI 인용 실험, GA4 AI Assistant 채널, 업체 검증 — 준이아빠블로그
- [3] ↑로고를 훔친 대행사, GEO라는 달콤한 용어 — 오즈의 지식토킹
- [4] ↑GEO 업체 선택 가이드: AI 검색 최적화 업체가 넘쳐나는 시대, 제대로 된 곳을 찾아야 하는 이유 — 지오랭크
- [5] ↑AI Features and Your Website — Google Search Central
- [6] ↑Do You Need an SEO? Tips for Hiring an SEO — Google Search Central
- [7] ↑New Research: AIs are highly inconsistent when recommending brands or products — SparkToro
- [8] ↑GEO: Generative Engine Optimization — arXiv
- [9] ↑디자이노블, MWC 2026서 AI 검색 가시성 플랫폼 BVI 출품 — AVING
Related documents
- Korea GEO Agency Comparison & Selection Guide 2026: Type, Engine Coverage, and ChecklistFive Korean GEO services categorized into four operational types: viral, content, measurement tool, and managed solution, and compared directly on engine count, Korean-language support, and pricing. Public specs for Zest Company, OPTIGEO, GPTO, SOHA, and BOIDA BVI plus a five-point pre-contract checklist, benchmarked as of 2026.
- GEO Company Evaluation Criteria: What to Look at Before You Choose (A Neutral Comparison)The decision criteria you need when picking a GEO vendor feels overwhelming. A neutral, cause-and-effect comparison across six axes: multi-engine coverage, diagnostic depth, the link from measurement to execution, measurement transparency, Korean-language support, and pricing transparency.
- GEO Agency Retainer Pricing in Korea 2026: Cost Bands, Service Scope, and What Drives the PriceA single monthly figure for a GEO agency retainer is misleading because measurement, diagnosis, and execution sit in separate pricing layers. This guide covers global monitoring tool list prices ($29: $499/month), Korean agency retainer bands (entry, mid-tier, full-service), the four cost drivers, and how to compare quotes on equal terms.
- GEO Cost & Contract FAQ: Price Tiers, Contract Types, DeliverablesAn FAQ for decision-makers wondering about 'GEO cost' and 'AEO pricing.' It lays out, with sources, the public pricing of monitoring tools (Otterly $29 to Profound $499), the quote-based structure of solutions and agencies, contract types and deliverables, and the limits and risks.
- GEO Agency vs. In-House vs. Solution: Comparing the Three Operating ModelsA decision guide for teams weighing 'GEO outsourcing vs. in-house.' A neutral comparison of three operating models: agency outsourcing, in-house ownership, and solution-based self-service, across structure, cost, speed, and fit.
- Why AI Visibility Tools Report Different Scores for the Same Brand: Causes and a Verification ProcedureWhy the same brand scores differently in every AI visibility tool, split into four layers (engine variance, collection design, metric definition, aggregation), with a side-by-side table of the published formulas from Profound, Peec AI, and Ahrefs, sample-size confidence interval thresholds, and a six-step verification procedure.
- GA4 AI Assistant Channel: Measuring ChatGPT, Gemini, and Claude TrafficGA4's AI Assistant channel: added to Default Channel Groups in May 2026, automatically classifies AI referral traffic, yet a structural blind spot remains: sessions that arrive without a referrer header. Here's what the channel captures, what it misses, and how to cover the gap.
- Best GEO/AEO Companies: Domestic & Global Agencies and Solutions Comparison Guide 2026A comprehensive answer to 'which GEO companies are worth recommending.' Compares domestic Korean (Intermajor, ZESTCOMPANY, Narr/Answer, BizSpring, etc.) and global monitoring tools, diagnosis solutions, and agencies by founding, headquarters, and differentiators.
- What Is GEO: The Definition of Generative Engine Optimization and How It Differs From SEOGEO (Generative Engine Optimization) is the strategy of getting your content cited in answers produced by generative engines like ChatGPT and Perplexity. Here is the definition, how it differs from SEO, and how it works.