The Neural Visibility Loop has four steps: SCAN (a baseline measured with an industry-specific question set), BUILD (bot access, server-rendered content, schema and entity consistency), WRITE (citable content and verifiable mentions) and MEASURE (the same questions re-asked regularly, at least monthly, across 6 engines). Each step’s output is the next step’s input; what we commit to is the work and the measurement, not an outcome.
What goes into each step, and what comes out?
| Step | Input | Output |
|---|---|---|
| 01 · SCAN | Interview with your sales and support teams, search data, competitor list | Question set, engine × question matrix, status report |
| 02 · BUILD | Technical findings from the status report | A prioritised technical task list and a dated change log |
| 03 · WRITE | Unanswered questions and the source map | New or restructured pages, mention work |
| 04 · MEASURE | The same question set | A report, monthly or more often depending on the plan, and the next period’s priority list |
How the steps run in practice depends on the service: the GEO agency page covers the monthly work and the Neural Visibility Audit page the one-off scan.
How is the question set built?
The question set is a sample of what your customers ask AI assistants. The baseline uses a core set of 30–150 questions depending on the industry; the number tracked afterwards depends on the plan. Questions are grouped by intent: discovery, comparison, purchase, verification. Once approved it stays fixed for the measurement; questions added later are tracked in a separate appendix so that months stay comparable.
For example questions by industry, see our industry files.
How is measurement done?
- Every question is asked in the 6 engines we monitor: ChatGPT, Gemini (including AI Overviews), Perplexity, Claude, Copilot and DeepSeek. Grok and Meta AI are covered by guides only.
- To reduce the effect of personalisation we use clean sessions where possible and repeat each question; the same question can get different answers at different times.
- Each answer is archived with its date, language and location; the brands named, their order and the cited URLs are recorded.
- The engines’ own reports are used too. Google Search Console’s generative AI performance report shows a site’s impressions in AI Overviews and AI Mode and has been available to all sites since 31 August 2026[1]. Bing Webmaster Tools’ AI Performance report shows how often your site is cited in AI answers on Copilot and Bing[2].
Which indicators are in the report?
| Indicator | What it tells you |
|---|---|
| Citation share | The share of answers to the question set that name or cite your brand |
| Recommendation position | Your place in answers that list several brands |
| Source coverage | How many of the cited URLs come from your site or from sites that describe you correctly |
| Sentiment score | A classification of the tone the answer uses about your brand |
| Wrong-facts list | False or outdated statements engines make about your brand, and their likely sources |
What don’t we promise?
- We don’t guarantee a place or a rank for a given question in a given engine; no agency can.
- We don’t produce paid links, fake reviews or artificial mentions.
- Engine behaviour changes; every finding in our reports carries a date.
- Our example scenarios are illustrative and never presented as real client results.
How does the loop change by industry?
The steps stay the same; the emphasis moves. In SaaS it’s comparison content and advertising rules, in hotels and travel multilingual pages and certificate details, in manufacturing technical product data, in retail branch data, in private education the licensed name and the programme and fee-announcement pages. Details in the industry files.
Frequently asked questions
01How long does one round of the loop take?
Scan frequency depends on the plan, from 4 scans a month to continuous scanning. Results are compared at least monthly and the next period’s priorities follow from them. The timeline for technical work is set after the baseline audit, based on scope; when change shows up in answers depends on the industry, the competition and the engine.
02Can we see the measurement data ourselves?
Yes. With the audit you receive the raw answer archive; for monthly plans the scope of the report and archive is written into the proposal. You can see, with dates, how each question was answered in each engine.
Sources
- Search Console Help — Generative AI performance report (Search) · accessed 28 September 2026
- Bing Webmaster Blog — Introducing AI Performance in Bing Webmaster Tools (Feb 2026) · accessed 28 September 2026