Skip to content
Rapillo Creative
Back to articles

Nir Levi's GEO and AEO Consulting: A Method Built on Recorded Evidence

A narrow but growing problem sits at the center of Nir Levi's consulting practice: a company may rank well in traditional search and still be described incorrectly, cited from the wrong source, or omitted entirely when a buyer asks an AI tool a direct comparison question. Levi, based in Haifa and working in search since 2008, treats this as an information and access problem rather than a ranking problem to be hacked. His service combines conventional SEO with Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO), but the more interesting part is the method behind those labels.

Where GEO and AEO fit into the search stack

The practical distinction on Levi's site is useful because the two terms often blur together in marketing. AEO work concentrates on the page itself. Is the answer near the top? Does a pricing explanation make sense when read alone? Is the product named clearly enough to survive outside its surrounding text? The AEO guidance on the site lists failure patterns: the answer arrives too late, the subject remains vague, features lack a use case, claims stay unverifiable, schema does not match the page, and essential content depends on client-side JavaScript that some crawlers do not render.

GEO, by contrast, extends outward to the sources AI systems rely on. It checks how a business is named, described and recommended across ChatGPT, Google AI Overviews, Gemini, Perplexity and Claude, and records which URLs those answers actually cite. This distinction matters because a mention without a company link, or a link to an old page, is a different problem from a missing mention. The two disciplines overlap, but separating them keeps a team from fixing the wrong layer.

How an engagement is scoped and run

Prospective work starts with business questions, target audiences, languages, and relevant AI systems. These are agreed before any work begins, along with price and timing. That scoping matters because a Hebrew answer and an English answer can draw from entirely different sources. A published research exercise on the site, covering 24 Perplexity answers collected on 11 October 2026, found an average source-host overlap of just 34.7 percent across paired Hebrew and English buying questions. The same-language repeat overlap was higher at 47.2 percent, but the point stands: language is not a cosmetic translation issue.

The work itself begins with a baseline. The consultant records the exact prompt, system, search mode, date and complete answer, then checks company information across the website and external profiles. Technical access gets the same attention. The public method distinguishes OpenAI's OAI-SearchBot, used for search, from GPTBot, which handles training, and applies similar separation to Anthropic's Claude bots and PerplexityBot. That distinction is important for organizations that want discovery without accidentally feeding specific pages to model training. The service page also says clients work directly with Nir Levi rather than through junior account managers, and engagements may take the form of fixed-scope audits, actionable roadmaps, or ongoing advisory sprints.

Deliverables designed for handoff

The output list is deliberately practical. Rather than a slide deck of vague opportunities, the work can produce a question inventory, an answer-and-source log, content and technical findings, a prioritized task list, Schema guidance, and content briefs or rewrites within the agreed scope.

Deliverable Practical function
Question inventory Groups category, comparison and branded questions by audience, language and platform.
Answer and source log Records full answers, sources and test conditions for later comparison.
Content and technical findings Connects each issue to supporting evidence and lists what still needs checking.
Prioritized task list Assigns affected pages, rationale, responsible role and a verification method.
Schema guidance Recommends markup types that match the content visible on the page.
Follow-up comparison Repeats the agreed questions under recorded conditions to compare changes.

What stands out here is the verification column. Rather than ending at "fix this page," the task list specifies who should handle the change and how to check whether it worked. That is a meaningful difference for marketing teams that have received plenty of audits but few implementable instructions.

Pricing: no published rate, by design

No consulting rates or fixed-price plans appear in the collected pages. The service page states that price and timing depend on the agreed scope, including the questions, languages, platforms and work involved. This is not a productized service; a company cannot buy a standard audit from a checkout screen. For some teams, that creates a longer sales process. For others, it is a signal that the work will not be forced into an off-the-shelf package.

Two free entry points exist: an AI visibility checker for ChatGPT and Gemini, and a no-signup worksheet for recording manual checks. The collected pages do not specify usage limits or technical requirements for either tool, so organizations planning heavy repeated use should ask about those details during scoping.

Who stands to benefit

The target audience is relatively specific: SaaS companies, startups and larger organizations, especially those with multilingual or multi-region sites. It also suits teams whose buyers compare products through AI tools before contacting sales. The bilingual Hebrew and English focus is a genuine differentiator for Israeli companies, because a strong English presence does not guarantee accurate Hebrew AI answers.

The public client list includes monday.com, ClickCease, Casino.com, 10Bet, MansionBet, Sensica and T·TIME. The About page describes engagement scope and technical focus for these clients, but avoids publishing confidential analytics. That restraint is consistent with the broader cautious measurement posture found throughout the site.

The measurement philosophy is the real product

Away from charts, Levi's most distinctive contribution may be how he refuses to conflate outcomes. The site separates whether a brand is mentioned, whether the answer links to its site, whether the brand is recommended, whether a visit arrives, and whether an inquiry results. A single changed AI answer is not accepted as proof of business impact. The 24-answer pilot illustrates the point: tracked brands appeared in 24 of 24 answers, but official-domain links appeared in only 22. That is not a failure. It is a distinction a less careful measurement would miss.

Signal What it records What it cannot prove
Mention Brand appears in an AI answer Citation, recommendation or traffic
Citation or link Answer points to a company page Anyone clicked the link
Recommendation Brand is presented as suitable for the need A customer inquiry or revenue
Accuracy Claims match a reliable source The change will propagate quickly

This level of rigor has practical consequences. Teams do not chase every AI fluctuation. They repeat the same questions under recorded conditions, keep the prompt stable, and investigate changes without assuming an edit caused every difference. Google's guidance that no special Schema is required for AI Overviews is also reflected in the recommended work: fix access and clarity first, then consider markup that matches the page.

Overall, the offering is best understood as evidence-led consulting for a moment when AI answers have made mixed-up or buried business information more visible and more costly. The trade-off is clear: no guarantees and no published fixed price, but a documented process, measurable baseline, and deliverables that can be handed to a marketing or development team.

Ready to work together?

+1 609-380-1994Call for a free consultation

Rapillo Creative

Marketing agency in Chesterfield, US

Business hours

Monday–Friday: 9 am–5 pm
Saturday–Sunday: Closed

Location

View Rapillo Creative on Google Maps