The seven steps, in order
The order matters. Steps one and two make you identifiable. Steps three and four make you quotable. Steps five and six make you credible. Step seven tells you whether any of it worked.
1. Lock the entity
One canonical Organization schema. One name, one description, one set of facts. AI engines cite brands they can identify without ambiguity.
A brand with three different descriptions across its site, LinkedIn and Crunchbase gets hedged out of answers. Collapse them into one, and the engines start naming the brand instead of the category.
In the audit: The audit checks whether your Organization schema is single, valid, and consistent with every external profile it can find.
2. Define your people
Founders and key roles as Person entities, linked to the Organization. Named humans dramatically increase citation confidence.
Two competing consultancies, same services. The one with named, linked founders gets cited by name. The anonymous one gets summarised as "some agencies".
In the audit: The audit checks whether your people exist as linked entities and whether their facts agree across the web.
3. Publish authority pages
Defined-term pages on each concept you want to own in your category. One deep page per concept beats fifty thin posts.
Fifty 600-word blog posts on overlapping topics compete with each other. Ten deep definition pages, one concept each, compound instead.
In the audit: The audit maps the concepts your category is asked about and shows which ones you have no page for.
4. Write atomic answers
Each section answers one question completely, so an AI engine can quote it without needing the surrounding context.
"As we discussed above, the third factor matters most" cannot be quoted. "Entity clarity is the fastest lever because engines stop guessing who you are" can.
In the audit: The audit scores whether your key pages front-load answers or bury them under narrative.
5. Align facts across the web
Same name, description, services, and location on LinkedIn, Crunchbase, directories, and press. Contradictions make engines hedge or skip you.
One directory listing an old service line or a retired founding date is enough to make an engine qualify its answer, or drop you for a competitor it is more certain about.
In the audit: The audit compares your on-site facts against the third-party sources engines actually read, and lists every contradiction.
6. Earn trusted references
Citations from sources AI engines already trust: major media, industry publications, Wikipedia, authoritative directories.
One mention in a recognised category publication moves citations more than fifty generic directory links, because the engine already trusts the source.
In the audit: The audit benchmarks your reference profile against the competitors currently being cited instead of you.
7. Measure monthly
Track citations across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews. Watch for regressions. Prioritise the next fix.
Visibility moves in both directions. Brands that measure monthly catch a drop in weeks. Brands that do not find out when their pipeline goes quiet.
In the audit: The audit is the baseline measurement: 25 dimensions, scored, benchmarked, with the gaps traced back to their cause.
The strategies behind the steps
The seven steps are the sequence. Underneath them sit six repeatable strategies that keep working after the first pass is done.
- •Entity engineering. One unambiguous Organization entity with linked Person, Service, and WebSite nodes.
- •Authority publishing. One indexable page per concept. Definition, FAQ, and explainer formats outperform blog rambles.
- •Atomic answer design. Front-load the answer. One question per section. Quote-ready prose.
- •Fact synchronisation. The same name, description, and services everywhere an engine can read them.
- •Trusted-reference acquisition. Citations from sources the engines already cite.
- •Monthly measurement. Track, catch regressions, prioritise the next fix.
Improving visibility versus increasing it
These are two different problems and they need different work. Improving visibility means fixing accuracy: the engines already mention you, but they describe you wrongly, hedge, or attach you to the wrong category. That is an entity and fact-alignment problem.
Increasing visibility means raising frequency: the engines rarely surface you at all for the questions your buyers ask. That is a coverage problem, solved with authority pages on the concepts you are missing and references that give the engines a reason to trust them. Most brands have some of both, which is why sequencing beats effort.
Google AI Overviews. A special case.
Google AI Overviews sit inside classic Google search. Unlike ChatGPT or Perplexity, they still lean on Google's traditional ranking systems as their first filter. If a page cannot rank in the top 10 for the underlying question, it will not surface inside the Overview either.
To improve visibility in Google AI Overviews specifically, do three things on top of the seven steps.
- •Earn the traditional ranking first. Overviews reward pages already visible in the standard results for the query.
- •Mark up the answer. FAQPage and HowTo schema make it trivial for Google to lift a quotable answer into the Overview.
- •Keep entity signals consistent. Google trusts the same Organization graph everywhere it sees your brand. Contradictions between your site, Google Business Profile, and third-party sources reduce trust.
How the engines differ
The fundamentals are shared, but the weighting is not. ChatGPT leans on training data plus retrieval, so consistent, long-standing facts matter most. Perplexity leans hard on live search, so freshly published, well-structured pages move fastest there. Claude is conservative and hedges when sources disagree, which makes fact alignment the deciding factor. Gemini and Google AI Overviews inherit Google's ranking layer first.
Optimise the fundamentals once, properly. Then tune for whichever engine your buyers actually use.
What does not work
- •Publishing AI-spun posts at scale. Volume without a distinct point of view gives the engines nothing worth quoting, and dilutes the pages that were working.
- •Keyword stuffing. Answer engines extract meaning, not term frequency. Repetition reads as noise.
- •Buying directory links. Untrusted sources do not transfer trust, and inconsistent listings actively damage entity clarity.
- •JavaScript-only pages. If the content is not in the served HTML, some crawlers never see it. Nothing else on this list matters if the page is invisible.
- •Duplicating a topic across several near-identical pages. The engines split their confidence and commit to none of them.