Nir Levi at a glance
- Role: Principal GEO, AEO & Technical SEO Consultant; Founder of Madrank Digital Ltd (Haifa, Israel) and Godrank.
- Experience: 18+ years in competitive search (since 2008) across B2B SaaS, cybersecurity and AdTech, digital health, and competitive international markets.
- Selected Client Experience: monday.com · ClickCease · Casino.com (UK, DE, FR, ES) · 10Bet · MansionBet · Sensica · T·TIME. See scope details on the About page.
- Works with: SaaS companies, startups, large organizations, and multilingual/multi-region websites.
- Working languages: Hebrew and English, with research, source mapping and roadmaps adapted to each market.
- Tools & Telemetry: Clean-session consumer UI sampling paired with DataForSEO LLM Mentions & LLM Responses, server-log/crawler verification (
OAI-SearchBot,PerplexityBot,Claude-SearchBot), Ahrefs, Google Search Console, and the AI visibility checker on this site. - Engagement Model: Direct work with Nir Levi—no junior account managers. Fixed-scope baseline audits, actionable team roadmaps, or ongoing advisory sprints.
- Contact: niro@nirlevi.com · LinkedIn.
The Casino.com engagement covered international SEO in 2018–2020 across the UK, Germany, France and Spain. Engagement scope and technical focus.
What GEO involves
Generative Engine Optimization (GEO) concerns how your company is cited, described, and recommended inside AI-generated answers across platforms like ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, and Claude. When a prospect asks an AI system to shortlist vendors, compare pricing, or evaluate technical fit, a search-enabled answer can draw on passages from your website and external sources. We examine the answer and the sources it actually cites.
The work covers three connected layers: technical access for search and retrieval crawlers, clear, extractable product facts and entity definitions on your site, and accurate representation across external sources such as review portals, comparisons, publications, and communities. SEO fundamentals remain relevant to Google’s AI search features. A top Google ranking does not guarantee a citation in a particular AI answer; each language and system must be verified directly.
Problems worth investigating
- Bilingual divergence between Hebrew and English (34.7% source overlap). English content and PR do not necessarily shape Hebrew AI answers—or vice versa. In the October 2026 study on this site, average cross-language source-host overlap was just 34.7% for paired buying questions. Each language and market needs its own buyer questions and source map. See the study and calculation.
- Blocked retrieval bots and JavaScript rendering gaps. A page that works in a browser may return initial HTML without its essential facts, or block requests at the server or WAF. We check server responses, initial and rendered HTML, and access rules by provider. Google renders JavaScript, but not all bots do; we do not assume every crawler sees the same page.
- Incomplete comparisons and missing buyer criteria on product pages. Before requesting a demo, buyers ask about ideal customer fit, pricing models, integrations, and limitations. If these facts are missing or buried in marketing copy, we check whether AI answers rely on third-party blogs or competitors, and identify the information your pages need.
- Inconsistent entity names and conflicting external facts. A rebrand, legacy subdomain, or outdated pricing tier can leave conflicting facts across your site, Schema markup, G2/Capterra profiles, Crunchbase, and press mentions. We identify discrepancies across owned and external sources and specify what needs updating.
- Unprotected domain and URL migrations. Changing URL structures, consolidating domains, or replatforming without checking previously cited pages can break source links in AI answers and external publications. Migration plans should explicitly audit those URLs and redirect them to appropriate destinations.
How we work together
We start with customer questions, examine answers and sources, then decide what your team should address.
1. Record the starting point (Live UI + API Telemetry)
Agree on buyer questions, target AI systems, and languages for your audience. We combine repeated clean-session checks in consumer interfaces, with test conditions recorded, and data from DataForSEO LLM Mentions and LLM Responses. Mention metrics, API-generated responses and answers observed in consumer interfaces are kept separate: these are different collection methods. Every question, answer, date, collection method, and cited URL is saved for comparison.
2. Identify the information buyers need
Break broad questions into useful research topics, such as price, integrations and support. These are research hypotheses; we call them searches executed by a system only when a record supports that claim.
3. Check company information
Review company and product names, descriptions and links across the website and relevant profiles. Check that Schema describes the same information readers see.
4. Review external sources
Examine review sites, comparisons and publications cited in answers. Recommend corrections to outdated information within each site’s rules. Publication and citation decisions remain with the external party.
5. Improve website answers
Write direct explanations with examples and sources where useful. Readers should understand the product and who it suits. See the AEO expert page for this part of the work.
6. Check technical access and HTML readability
Verify robots.txt, server/WAF responses, and server-rendered HTML so essential content does not depend only on client-side JavaScript. Distinguish search and retrieval from training: OpenAI uses OAI-SearchBot for search and GPTBot for training; Anthropic separates Claude-SearchBot for search, Claude-User for user-directed retrieval, and ClaudeBot for training. PerplexityBot serves search. Where provider verification information is available, we cross-check requests against it instead of relying on a User-Agent name alone. We compare access findings with the live-interface checks and DataForSEO observations recorded in the baseline to define what the team should investigate or fix.
7. Repeat the checks
Repeat questions under recorded conditions and compare mentions, links and factual accuracy. Agree on the frequency and support before work begins.
What you receive
- A baseline record: questions, answers, systems and sources checked.
- Prioritized tasks: what to change, where, why and who will do it.
- Information and Schema guidance: consistent descriptions of the company and products.
- External-source review: facts to update and where they appear.
- Content briefs or page rewrites: according to the agreed scope.
- Follow-up comparison: where included, record what changed and what did not.
How we assess progress
Track three things separately: whether the company is mentioned, whether the answer links to your site and whether the information is accurate. Also calculate the proportion of collected answers that mention the company, by system and language. This describes the sample, not the company’s market share.
Visits and customer inquiries are separate measures. A citation does not establish that anyone clicked it, and a click does not establish that an inquiry arrived.
Check your AI visibility or start tracking with your team
Before scoping a full engagement, you can use the free working tools on this site to run initial checks or organize your team's observations:
- AI Visibility Checker: Check mentions and cited source links in AI answers within the tool's defined scope.
- AI Visibility Worksheet: A structured working template to build a buyer question set, separate brand mentions from linked citations, and log repeated observations over time.
An inspectable record: Hebrew and English answers
The bilingual study published on this site collected 24 live Perplexity answers across six commercial buying questions on 11 October 2026, tested twice per language in Hebrew and English. Data collection and calculations used Codex assistance; Nir Levi's analysis and interpretation appear separately in the report.
Each tracked category brand appeared in all four answers for its category, while average cross-language source-host overlap was only 34.7%. The measure compares shared source hosts with the combined host set in each Hebrew/English answer pair, then averages across 12 pairs. I recommend auditing each target language separately and retaining full answer logs and cited URLs to identify what information needs updating on your website and across external sources.
Read the full study · Download research observations (CSV) · Download the answers and calculations (ZIP).
Methodology note: This published dataset is a dated, single-day Perplexity panel with two rounds per language. It demonstrates how cross-language citation divergence is measured; its findings do not establish what happens across all questions or systems. In client engagements, we adapt this transparent logging method to your buyer questions, target markets, and agreed AI systems.
Who this work suits
SaaS companies, startups and enterprises that want to understand how AI describes them, with teams able to update the website after the review. The scope can cover one language or several markets.
If important pages are blocked or the team cannot currently edit the site, start there. Priorities should reflect what can be implemented.
Frequently asked questions
What does a GEO expert in Israel do?
Examines how a company appears in AI answers, which sources support those answers and what information needs improvement. Work can cover Hebrew and English according to the audience and market.
Is GEO the same as SEO?
They overlap in content and technical work. SEO measures include impressions, rankings and search visits. GEO also examines mentions, citations and factual accuracy in AI answers.
Which AI systems are covered?
We agree on relevant systems in advance, such as ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity or Claude. Each is checked separately with its test conditions recorded.
How do you measure AI visibility?
Save questions and answers, count mentions and links separately and check factual accuracy. Start with the AI visibility checker; a broader assessment can include further questions and repetitions.
How long until answers change?
There is no dependable date for AI answers to change. We agree on a work schedule and follow-up checks, then examine what happened after the changes were published.
Can you guarantee a ChatGPT citation?
No. I can commit to the agreed work: checks, documentation and clear recommendations. The system decides which sources to cite.
Can you work on a site serving other countries?
Yes. I work in Hebrew and English, choosing questions appropriate to each market and audience.
What is the difference between GEO and AEO?
The terms overlap. My AEO work focuses on questions and answers within pages; GEO work also examines external sources and broader visibility. See the AEO expert page.
How do we start?
Send your website and the question you want to investigate. We agree on systems, questions and deliverables, then use the findings to decide whether further work is useful.
We already have an in-house SEO or content team. How do we work together?
Having an in-house team is an advantage. Most internal teams are busy executing ongoing roadmaps and need a clear diagnostic baseline, a map of which external sources AI answers rely on, and concrete specifications for content and development. We agree in advance on the division of work: I provide the investigation, prioritized task list, Schema/technical specs, and content briefs, and review the changes with your team once implemented.
Should we buy an AI monitoring tool or work with a consultant?
A monitoring tool can record whether your name appeared in a prompt set over time. A mention count alone does not explain why a crawler could not read a page, which product facts are missing from a comparison page, or how to improve the content and Schema. We use monitoring data where helpful, and connect it to investigation that turns findings into tasks your developers and writers can ship.
Do we have to commit to a monthly retainer?
No. Most projects begin with a focused baseline review and action plan, with scope and budget agreed in advance for your website and markets. Once your team receives the findings and task list, we decide together whether ongoing advisory support or periodic follow-up checks have a business case.
Talk about your website
Send your website and a few words about your customers and what you want to improve. We can work out where to start.
Let’s talk · niro@nirlevi.com · LinkedIn
