The biggest red flag is any guarantee. Nobody outside OpenAI, Google or Perplexity controls what their systems say, so an agency promising guaranteed citations, mentions or "top spots in AI answers" is selling certainty that does not exist. Good GEO (generative engine optimisation) proposals describe methods, evidence and measurement, and they are honest about variability.
Key Takeaways
- Guarantees of visibility in AI answers are the clearest warning sign. AI outputs vary from one response to the next.
- Ask for the method in plain language. "Proprietary" is not an answer.
- Look for SEO foundations, content, authority-building and business-outcome reporting, not a single tactic.
- Dashboards are measurement. A proposal whose deliverable is "access to a dashboard" is a tool resale.
- Demand case studies with baselines, time frames and what else changed.
- This list applies to me too. I have written it so you can hold any agency, including mine, to it.
Why GEO proposals need a separate checklist
GEO, AEO and "AI SEO" are new enough that buyers cannot easily tell competent from opportunistic. Traditional SEO has years of shared vocabulary and reference points. AI search does not yet. That gap is where overpromising thrives.
Two pieces of evidence set the context. SparkToro's research (600 volunteers, 2,961 responses across ChatGPT, Claude and Google AI, run in November and December 2025) found less than a 1 in 100 chance that an AI gives the same list of brands in any two responses to the same prompt. You can read the SparkToro findings. And several agencies themselves say guarantees are a problem: SEO.com's AI SEO agency red flags says you can appear in AI results for one searcher and not the next, and WebFX's guide to choosing a GEO agency lists guaranteed outcomes as its first red flag. Note that both are agencies with their own services to sell, so read them as informed opinion, not neutral research.
The 9 red flags
1. Guaranteed citations, mentions or rankings
This is the headline. An agency can influence the inputs (content, authority, structure) but cannot control the outputs of third-party AI systems. Acceptable language sounds like "we aim to increase mention rate across a defined prompt set, measured over 90 days." Unacceptable language sounds like "we guarantee you will appear in ChatGPT."
2. "Rank #1 in AI" language
Position in an AI answer is unstable. SparkToro called ranking position in AI outputs unreliable and favoured visibility percentage across many runs. If a report shows "you are #2 in ChatGPT" from one query, ask how many samples sit behind it.
3. Secret or vague methods
"Proprietary AI framework" with no explanation usually means there is either nothing there or something you would not approve of. Ask them to walk through what they would change on your site in month one.
4. One tactic sold as the whole strategy
llms.txt files, FAQ schema, or a single directory push are sometimes pitched as the answer. SE Ranking's analysis of 129,000 domains and about 216,000 pages reported that llms.txt had negligible effect and that FAQ-schema pages did not outperform those without (both correlational findings from one vendor's study; see their write-up). Good GEO is a bundle: technical accessibility, content quality, authority signals, and third-party presence.
5. Ignoring SEO fundamentals
If a proposal skips crawlability, indexation, site architecture and content quality, it is building on sand. AI systems that retrieve from the web depend on pages that are findable and trustworthy.
6. Visibility-only reporting
Mentions and citations are leading indicators. If nothing connects them to leads, sales or pipeline, you cannot tell whether the spend is working. Ask for a reporting sample with both.
7. No case studies, or case studies with no baseline
Evidence should show a starting point, a time frame, actions taken and what else changed (a rebrand, a press hit, a seasonal spike). "Visibility up 300%" without a denominator means little.
8. Tool resale dressed as service
Some proposals are mostly a monitoring subscription plus a monthly PDF. Tools are fine, but you can often buy them directly. Check what human work sits on top: strategy, writing, editing, outreach.
9. Pressure and scarcity
"Competitors are already ahead, you must sign this week." AI search is moving, but not so fast that a two-week diligence process costs you the market.
Questions to ask, and what a good answer sounds like
| Question | A good answer | A bad answer |
|---|---|---|
| How do you define success? | Prompt-set mention rate plus qualified traffic and pipeline, with a baseline | "Top rankings in AI" |
| What will you change in the first 30 days? | Specific pages, technical fixes, content briefs | "Optimise for AI algorithms" |
| Can you guarantee citations? | "No. Here is what we control and what we don't." | "Yes" |
| Which pages do you prioritise? | High-intent service, pricing, comparison and product pages | Blog posts across the board |
| How do you measure? | Repeated runs across engines, percentage visibility, raw response samples | Single-query screenshots |
| What have you tested that failed? | A candid example | "Everything works" |
| Who does the work? | Named people with roles | "Our AI" |
What a credible GEO scope contains
If I were evaluating a proposal, I would want to see most of these:
- Audit: crawl and index status, bot access rules, page speed, structured data accuracy, content gaps against buyer questions.
- Prompt research: a defined list of prompts by intent (category, comparison, problem, brand), agreed with you.
- Baseline measurement: repeated sampling across the engines your buyers use.
- Content work: answer-first pages, original data or experience, clear sections, updated facts, real authorship.
- Authority work: reviews, credible mentions, partnerships, digital PR, and presence in the communities where your buyers ask questions.
- Reporting: monthly mention rates, cited sources, and traffic or pipeline tie-ins, with caveats about variability.
- Exit terms: you own the content and accounts; monthly or quarterly contracts where possible.
A note on honesty and my own bias
I am an organic growth and content strategist, so I have an interest in this market. I am telling you not to trust guarantees because I cannot make them either. What I can do is describe the process and show past work. The Masai School programme I contributed to (Instagram from 26K to 117K followers, LinkedIn from 50K to 200K) was an organic social and content project, not a GEO guarantee, and I would not claim otherwise. It shows the kind of evidence you should ask any agency to show: a starting figure, an end figure and a time frame.
How to run your own two-week diligence
- Send the same brief to three agencies.
- Ask each to explain their method on a call, and record what they commit to.
- Request two references you can contact, and one example of a failed test.
- Compare scopes using the table above.
- Start with a paid audit or a three-month pilot, not a 12-month contract.
- Set success metrics in writing, including what happens if they are not met.
FAQ
What is GEO?
Generative engine optimisation is the practice of making your content and brand easier for AI answer engines to find, understand and cite. It overlaps heavily with SEO and PR.
Can any agency guarantee I will appear in ChatGPT?
No. Outputs vary between responses and depend on systems the agency does not control. Treat guarantees as disqualifying.
How long does GEO take to show results?
It varies by site authority and competition. Treat any promise of results in days or a few weeks with caution, and agree on a baseline and review period, such as 90 days.
Is GEO different from SEO?
Partly. The foundations are shared: crawlable pages, quality content and authority. GEO adds attention to prompts, third-party mentions and how answer engines summarise your content.
What should a GEO report include?
Visibility percentage across a defined prompt set over repeated runs, cited sources, accuracy of what AI says about you, and links to traffic, leads or revenue where measurable.
Are AI visibility tools enough on their own?
No. They measure. Someone still needs to do the content, technical and authority work.
How much should GEO cost?
Prices vary widely by market and scope, and I have not verified a benchmark to quote. Compare scopes and deliverables rather than headline price, and start with a small paid pilot.
Should I sign a long contract?
Prefer short terms or a pilot until the agency has shown measurable progress. Ensure you own all content and accounts if you leave.
What if an agency's case studies look impressive?
Ask for the baseline, time frame, actions and other factors that changed. Verify by contacting a reference.
Work with me
I am Younus Fardeen, an India-based organic growth, SEO and content strategist with 4+ years of marketing experience across edtech and startups. If you would like a second opinion on a proposal or a plain-spoken plan of your own, see my work and reach out through the contact form at younusfardeen.in.