MaxAEO Review

A decade of SEO rested on a comforting property: you could see the result. Type the query, look at the page, know where you stood. AI assistants removed that. A buyer asks ChatGPT which tool to use, gets three recommendations and a summary of each, and acts on it — and the brands involved have no idea the conversation happened.

MaxAEO, operated by HIII PTE. LTD., is one of a new category built around that blindness. It runs queries across eight AI engines, records whether you were mentioned, how you were described, who was recommended instead, and which sources the answer leaned on.

This review covers what these metrics genuinely tell you, why citation tracing is the only part that leads to action, how much influence anyone actually has over model output, and whether this belongs in your stack yet.

Answer Engine Optimisation, Explained Quickly

• The problem is real: buyers increasingly ask an assistant rather than a search engine, and you never see what it said about you.
• Eight engines tracked — ChatGPT, Gemini, Perplexity, Claude, Copilot, Grok, Google AI Mode and Google AI Overview.
• Mention rate is the headline metric: how often you appear at all when a relevant question is asked.
• Citation tracing is the actionable part, because it shows which sources the models drew on — and those are things you can influence.
• Measurement is solid; control is not. Nobody can guarantee what a model will say, and any vendor implying otherwise is overselling.

The Vocabulary Of This New Category

AEO has arrived with its own jargon. These are the terms that matter.

AEO

Answer Engine Optimisation — influencing how AI assistants describe you, as opposed to where you rank in a list of links. The shift is from position to portrayal.

Mention rate

How often your brand appears at all across a set of relevant prompts. It is the field's equivalent of impressions, and the first number anyone looks at.

Citation tracing

Identifying which sources an answer drew on. The most actionable metric here, because sources are things you can influence and model weights are not.

Prompt research

Working out which questions your buyers actually ask an assistant. The equivalent of keyword research, and just as foundational — measure the wrong prompts and every downstream number is meaningless.

Sentiment

Whether the description of you is favourable, neutral or unfavourable. Being mentioned badly is worse than not being mentioned, which is why mention rate alone is a poor scorecard.

What Eight Engines Buys You

The tracked list — ChatGPT, Gemini, Perplexity, Claude, Copilot, Grok, Google AI Mode and Google AI Overview — is broad, and breadth matters more here than it did in search.

In classic SEO, one engine dominated and optimising for it covered almost everyone. AI assistants have no such consolidation. The models have different training data, different retrieval behaviour and different tendencies, and a brand can be well represented in one and invisible in another for reasons that are not obvious from outside.

Splitting Google AI Mode from AI Overview is a sensible distinction rather than padding, since they are different surfaces with different behaviour.

The practical value of tracking all eight is finding the asymmetries. Strong in Perplexity and absent from ChatGPT is a specific, investigable finding — usually about which sources each one reaches for — and it is invisible if you only check the assistant you personally use.

Which Metrics Actually Lead Anywhere

The feature set covers AI Visibility Overview, Mention Rate Analysis, Competitor Benchmarking, Sentiment Analysis, Citation Tracing, Prompt Research and Content Optimization. They are not equally useful, and it is worth being blunt about the ranking.

Citation Tracing Is The One That Matters

You cannot edit a model's weights. You can influence what it retrieves, and citation tracing tells you what that currently is.

If answers about your category consistently cite three comparison sites, a documentation page and a Reddit thread, you have a concrete list of places to be accurate, present and well described. That is a real workplan, which is more than most metrics in this category produce.

Prompt Research Sets Everything Else

Every other number depends on measuring the right questions. Tracking prompts nobody asks produces a dashboard full of meaningless trends.

This is the same discipline as keyword research and it is harder, because conversational questions are longer, more varied and less observable than search queries. Expect to iterate on the prompt set rather than defining it once.

Sentiment And Benchmarking As Context

Sentiment matters because appearing in an answer is not automatically good — being described as the cheap option with poor support is a worse outcome than absence.

Competitor benchmarking supplies the only meaningful baseline. A 40% mention rate means nothing alone; 40% against a competitor's 75% on the same prompts is a finding you can take to a meeting.

The Honest Limits Of Influence

This is where the category needs scepticism, and where a review earns its place.

AI answers are generated from model weights and retrieved sources. You cannot touch the weights. You can influence retrieval by being present and accurately described in the sources models pull from, which is genuinely useful — but the causal chain from 'we improved our documentation' to 'ChatGPT now recommends us' is long, indirect and largely unobservable.

Results also change without you doing anything. A model update can move your mention rate overnight for reasons no vendor can explain, because the vendors cannot see inside the models either. Any tool in this category is measuring a system whose behaviour it does not control and cannot fully account for.

The measurement is worth having anyway — knowing you are invisible is better than not knowing. But treat AEO as a monitoring discipline with weak levers rather than a channel you can dial up, and be wary of anyone in this space promising outcomes.

You can measure what the models say. You can influence what they read. Nobody can promise what they will conclude — and a vendor who does is describing a product that does not exist yet.

Making It Useful Rather Than Decorative

The failure mode with visibility tools is a dashboard everyone glances at and nobody acts on.

1. Build the prompt set from real buyer questions — sales calls and support tickets, not a brainstorm.
2. Establish a baseline across all eight engines before changing anything, so you can tell a real movement from noise.
3. Read sentiment alongside mention rate from the start. Rising visibility with unfavourable framing is a worse position than where you began.
4. Take citation tracing seriously: list the sources appearing repeatedly, and check every one for accuracy about you.
5. Fix factual errors in third-party sources first. It is the cheapest available action and the one most likely to change an answer.
6. Benchmark against two or three named competitors on identical prompts, since absolute numbers mean little.
7. Review monthly rather than daily, and expect model updates to produce swings you cannot attribute to anything you did.

Coverage, Metrics And Operator