Most advice on this subject is written for a company at one particular stage and then applied by companies at four other stages, which is why so much of it lands badly. Somebody with a crawler access problem does not need content strategy. Somebody being actively displaced by a competitor does not need an introduction to schema.
So before absorbing any recommendation, identify where you actually sit. The stages below are sequential, each one has a distinct bottleneck, and the whole point of improving AI Search Visibility is to move up one level at a time rather than skipping ahead.
Stage Zero: Structurally Invisible
What it looks like: assistants cannot describe your company at all, or describe it wrongly. Zero appearances across every query you test.
The bottleneck is almost always mechanical. Crawler directives blocking the bots that matter, filtering at the CDN level, a site built so that content only appears after script execution, or pages that exist but have never been recrawled.
Companies at this stage frequently do not know it. Nothing in conventional reporting flags it, and rankings can be perfectly healthy while this is true.
What moves you up: technical access, verified rather than assumed. Test whether your pages can actually be read as plain text. This is a developer task measured in days, not a marketing programme.
Do not spend anything on content while you are here. It is the one stage where additional investment produces literally nothing.
Stage One: Present but Unrecommended
What it looks like: assistants can describe your company but never bring it up unprompted. Ask about you directly and you get an answer. Ask a buying question in your category and you are absent.
This is the most populated stage, and the most misdiagnosed. Teams here assume they need authority, when what they usually need is definition.
The bottleneck is clarity of what you are. If your pages cannot support a flat statement of what you sell and to whom, a model has nothing to offer when a buyer asks who to consider. Vague positioning, differentiators expressed as adjectives, and service descriptions that drift between pages all produce the same result: recognized, not recommended.
Ken Ganley Mentor CDJR in Ohio sat here despite being a substantial local business. The dealership was known offline and had barely any page level presence in AI answers. The work that shifted it was not brand building. It was making specific, concrete inventory and service information available in a form that could be extracted and used, which produced sharp mention growth inside two months.
What moves you up: rewriting existing commercial pages so the facts sit at the front, and reconciling how your business is described across every property you control.
Stage Two: Occasionally Recommended, Inconsistently
What it looks like: you appear sometimes, unpredictably, usually on lower competition informational questions rather than buying questions.
Progress is real here and it feels frustrating, because results fluctuate week to week and the pattern seems random.
The bottleneck is corroboration. Your own site now says clear things about you. Nothing else does. Models weight independent confirmation more heavily than self description, so your inclusion depends on how much competing material happens to surface for a given phrasing.
This is where self managed programmes stall most often, because the remaining work is outreach rather than editing. Directory accuracy, association listings, partner pages, third party coverage, knowledge panel correctness. None of it is intellectually demanding and all of it takes sustained hours.
What moves you up: external sources confirming the claims your site makes. Start with correcting the wrong ones, which is faster than creating new ones.
Stage Three: Consistently Recommended in a Narrow Band
What it looks like: reliable presence on the questions you have worked on, absence outside them. Real commercial benefit, clearly bounded.
Companies here often plateau for a year without realizing they have plateaued, because their reporting only covers the queries they already win.
The bottleneck is coverage of adjacent questions. Buyers do not ask one question. They ask a sequence, and the earlier questions in that sequence are frequently owned by somebody else who then shapes the whole consideration set. Winning the final comparison question matters less than being present when the buyer is still working out what they need.
Platform concentration also becomes a limiting factor at this stage. Results genuinely diverge by platform, and a company strong in one place can be entirely absent in another where a meaningful share of its buyers research. Something like track ai rankings on perplexity exists precisely because a single combined figure will not reveal that gap.
What moves you up: mapping the full research sequence rather than the purchase decision, and auditing platform by platform rather than in aggregate.
Stage Four: Broadly Established and Under Pressure
What it looks like: strong presence across a wide question set, and periodic unexplained losses as competitors move.
The work changes character completely here. It stops being construction and becomes defense, which most organizations are structurally bad at because defense has no launch date and produces no announcements.
The bottleneck is response speed. Positions shift when competitors publish and when platforms revise how they select sources. If nobody notices for a quarter, ground gets lost that takes two quarters to recover.
EC Council, the global certification body behind CEH and CND, illustrates the scale involved at this level. Already globally recognized with a large content estate, the programme ran six months rather than two, working through technical page restructuring and schema across a substantial body of educational material. Large estates take longer and yield more, and they also require more to maintain.
What moves you up: nothing. There is no stage five. This is the steady state, and the objective becomes staying here.
Diagnosing Your Own Stage Accurately
Two mistakes are common and both are expensive.
Assuming you are further along than you are. Teams see occasional mentions and conclude they are at Stage Three when they are at Stage Two with lucky weeks. The test is consistency across repeated runs of the same query, not whether you have ever appeared.
Assuming you are further behind than you are. Companies with genuine authority sometimes conclude they need to build from nothing, when in fact they are at Stage One and the fix is a fortnight of restructuring. Existing conventional search strength usually means the gap is presentation rather than substance.
Run each of your main buying questions three times across a fortnight. Consistent absence means you are below Stage Two. Inconsistent presence means Stage Two. Consistent presence on some questions and none on others means Stage Three, and that gap is where your next quarter should go.
Why Skipping Stages Fails
The sequence is not arbitrary. Each stage’s fix depends on the previous one holding.
Content produced at Stage Zero is never read. Corroboration built at Stage One confirms a description that is still unclear. Coverage expansion at Stage Two spreads thin material wider. Defense at Stage Three protects a position that was never broad enough to matter.
Which is why the most common expensive mistake is buying a Stage Three programme while sitting at Stage One. The deliverables all arrive, everything gets implemented, and the results do not come, because the work assumed a foundation that was not there.
Identify the stage first. The correct next action follows from it almost automatically, and it is usually smaller and cheaper than whatever you were about to commission.
