AI visibility: get ChatGPT to name your brand - neuroflash

AI visibility · Digital Twins · A package for your agent

Get recommended in ChatGPT for the questions your customers actually type.

An AI answer names 2.8 brands. Yours makes the cut when your content matches the real questions: 3.27× the odds of being cited.

This package gives Claude Code, Codex and other agents the skills to do it: generate real user questions with neuroflash Digital Twins, deliver content that answers them, measure the results.

Book a free intro call Run a free quick scan See the package
neuroflash Digital Twins, AI personas built on real survey data
CITED BY AI IN 3 MONTHS GUARANTEE*

Without

  • You guess what customers ask ChatGPT
  • Customers can’t find you in ChatGPT
  • Hours of work or big invoices for articles, posts, pages and video

With your agent

  • Digital Twins supply the real questions your audience asks
  • Customers reach your website from ChatGPT
  • 90% automated: articles, posts, offer pages, newsletters, video editing. Same effort, roughly 10× the output

3.7% → 26%

ChatGPT mention rate for jonathanmall.com in nine weeks

3.27×

Odds of being cited when your content semantically matches the real question

85 to 98%

Digital Twin accuracy, built on over 1 million real survey profiles

The difference: real questions instead of guessed prompts

AI answers cite the page that fits the question. Yet nobody knows which questions people actually type. Digital Twins do.

Queries from a standard LLM

Length
15 to 25 words, complete sentences
Style
Perfect grammar and punctuation
Data source
None. Assumptions about your audience

Queries from neuroflash Digital Twins

Length
Median of 11 words, calibrated to the real distribution
Style
46% start in lower case, 12% are bare keyword fragments
Data source
Over 1 million real survey profiles since 2017, 85 to 98% accuracy

Distribution measured on around 20 million real user queries on our platform.

Twin avatar Alina, 25, advertising and marketing, Berlin

Alina, 25

Advertising and marketing, Berlin

ai speaker who gets gen z audience

7 words, all lower case

Twin avatar Nadine, 37, HR assistant, Stuttgart

Nadine, 37

HR assistant, Stuttgart

ai keynote cost leadership event

5 words, a bare keyword fragment

Twin avatar Korbinian, 63, head of sales, Rosenheim

Korbinian, 63

Head of sales, Rosenheim

is an ai keynote worth it for a mid sized company

11 words, a fully phrased question

Illustrative example twins. The avatars and personas come from our Digital Twin system; the queries are examples of the process.

Your agent calls the twins directly through the neuroflash MCP. Define the audience, generate the query set, freeze it, measure.

What you get: the package

A package of skills and processes that your agent runs, plus access to neuroflash Digital Twins via MCP. No course to read through. Working instructions your agent understands: audience, query set, measurement, content rules, deployment and re-measurement.

Illustration: a figure in a crowd stands in the spotlight cast by a chat bubble

You get seen when your customers ask

ChatGPT, Claude, Perplexity and Gemini name you because your page answers the questions your audience really asks. The Digital Twins supply the questions; the package supplies the structure.

Illustration: connected tiles forming a closed loop

Everything connected, in one process

Audience, query set, measurement, articles, LinkedIn posts, images, video, deployment and re-measurement all mesh together. Your agent runs the loop every week, with no tools for you to stitch together.

Illustration: a smartphone showing three new appointment notifications

More enquiries, no agency

Brands that get named in AI answers receive enquiries from people who are already convinced. The weekly measurement shows your progress in numbers. The clients you win cover the cost of the package.

  • Digital Twin access via MCP

    Simulate your audience, generate realistic queries, test content before it goes live.

  • Website migration playbook

    WordPress to static, in 25 lessons.

  • Website editor with rollback

    Publish changes and undo them with one click.

  • GEO article writer

    Answer-first, so the answer sits in the first 60 words.

  • AI visibility audit

    A monthly query matrix across four engines.

  • LinkedIn and relationship engine

    Reach and contacts wired into the same loop.

  • Copywriting system

    Your voice, repeatable, without sounding templated.

  • Image generation

    Visuals for articles, social and pages, in your own look.

  • Video repurposing

    One recording becomes clips, pages and posts.

  • 7-day integration programme

    Daily documents your agent uses to set up the package in one week.

How the handover works

1

You get the package

Skill files, process documents and the 7-day programme.

2

Your agent runs it

Connected to neuroflash Digital Twins via MCP. Claude Code, OpenAI Codex, Hermes and others, on your own machine.

3

You measure every week

The same frozen query set, progress in numbers.

Why this problem is new

Your rankings can hold steady while your brand disappears from every AI answer. Three findings explain why.

Search has moved

Your customers ask ChatGPT, Gemini, Perplexity and Copilot before they open Google. An answer names around 2.8 brands. Everyone else drops out of the decision.

Classic SEO falls short

In our study, backlinks showed no independent effect (p = 0.42 and p = 0.15). FAQ schema on its own: no reliable effect (p = 0.27). Both remain hygiene; neither is a lever.

The strongest lever is fit

Semantic fit between the real question and the text passage: 3.27× the odds of being cited per standard deviation (95% interval 2.68 to 4.01). Answer within the first 60 words: 2.02×. Title using the words of the question: 1.96×. List and table density: 0.89×.

Fit requires real questions

Real users type a median of 11 words; 46% start in lower case and 12% are bare keyword fragments that a language model would never produce. Measured on a sample of around 20 million real queries on our platform. This is exactly the distribution our Digital Twins deliver, because they are calibrated on real behaviour.

The process inside the package: the weekly 7-step loop

No project with an end date. A cycle that runs once a week, with results you can read on day seven.

1 Simulate audience 2 Real queries 3 Measure 4 Read the gaps 5 Write content 6 Publish 7 Re-measure

Powered by your agent

Your coding agent runs the loop

Your coding agent

Claude Code · OpenAI Codex · Hermes · and others

Your coding agent runs the loop: Claude Code, OpenAI Codex, Hermes and others.

  1. 1

    Simulate your audience

    Digital Twins

    Digital Twins model the people who decide whether to choose you.

  2. 2

    Generate queries that sound human

    Digital Twins

    Calibrated to the real distribution and locked in as a frozen set.

  3. 3

    Measure

    Across ChatGPT, Gemini, Copilot, Perplexity and Claude, with Rankscale and Peec.ai.

  4. 4

    Read the gaps

    Branded and unbranded kept apart, because the two gaps have different causes.

  5. 5

    Write twin-tested content

    Digital Twins

    Answer-first, fact-checked before sign-off.

  6. 6

    Publish

    Static, fast, with one-click rollback.

  7. 7

    Re-measure after seven days

    The same frozen set, so the numbers stay comparable.

A twin query and matching content merge in phase into a golden record; an LLM query and mismatched content produce a cracked record
Content in phase with real questions gets cited. Generated noise on the wrong frequency ends up in the bin.

What you can expect

15 to 30%

The ChatGPT mention rate this process reaches on your own query set, in our experience

On jonathanmall.com the mention rate rose from 3.7% to 26% in nine weeks, meaning 26 out of 100 answers in the top-100 query set. A mid-sized European B2B company went from 3% to around 60% citation share with the same loop.

Your results depend on the competition in your niche. The first to give structured answers to their customers’ real questions wins faster. In crowded categories it takes longer, and results may land below this range.

A benchmark from our runs so far, not a guarantee. The guarantee for the Cohort and Done-with-you packages is set out under pricing.

Four case studies

One personal website, one large corporation, one mid-sized B2B company, one brand-new wiki. Four very different scales, the same loop.

Case A

jonathanmall.com

The website of our CIO Dr. Jonathan Mall, keynote speaker. Our own test bed for the loop.

3.7% 26%
ChatGPT mention rate, 2 July to 5 September 2026, measured with Rankscale.

Starting point

Spring 2026: around 50 visitors a month. WordPress with 468 KB of homepage HTML and 60 scripts, PageSpeed 70, roughly 16 Google impressions a day. ChatGPT mention rate of 3.7% on the frozen query set. Completely absent for “best AI keynote speaker”.

What we did

Out with WordPress, in with a static site. A weekly 7-step loop, 44 review-checked articles in the backlog, no more than two new pieces published per week, plus video and LinkedIn repurposing.

Result

PageSpeed from 70 to 97, homepage HTML from 468 KB to 14 KB. Google impressions from about 16 to about 287 a day, clicks from under 1 to around 2 a day. ChatGPT mention rate from 3.7% to 26%, meaning 26 out of 100 answers in the top-100 query set, nine weeks after the first measurement (measured on 5 September 2026). Three keynote enquiries through the optimised speaker pages, two of them booked, around €10,000 in revenue. One client asked ChatGPT for the best AI keynote speaker in Hamburg, got our CIO’s name, checked the website and booked a keynote in the mid four figures.

Status: a single website, measured honestly, not a controlled sample.

“I just booked my second gig from an organiser who found me through ChatGPT.”

Participant testimonial

Case B

EY

Employer branding, together with Ketchum and Rankscale.

Starting point

Top talent asks AI chatbots instead of Google. EY was barely visible in those answers.

What we did

Ketchum provided the strategy and briefing. Using our Digital Twins, grounded in over 1.8 million real platform users, we simulated four audiences across four journey phases: school students aged 16 to 19, university students aged 19 to 27, young professionals aged 23 to 32, professionals 32+. Rankscale validated and tracked visibility on ChatGPT, Gemini, Google AI Mode and Perplexity.

Result

1,911 realistic AI queries as the foundation of the content strategy. One insight from them: real users type a median of 7 words, without punctuation, in keyword fragments. Standard AI produces 15 to 25 words in perfect grammar. Ketchum took on content strategy and seeding, based on the gaps per audience and engine.

Status: enterprise project with an agency partner. The figures come from the public case study.

Read the EY case study

Case C

Mid-sized B2B company from the EU

The B2B study from our webinar with Rankscale: a mid-sized European company, the same loop, on a smaller scale. Anonymised.

Starting point

A mid-sized B2B company with its own domain and existing expert content. On a fixed set of buyer prompts, its own website was used as a source in only 3% of AI answers.

What we did

The same loop as for EY, on a smaller scale: existing content instead of new campaigns. The existing pages were rebuilt with a clear structure, clean entity naming and answer-shaped paragraphs, backed by a steady flow of expert content on the company’s own domain. Measured before and after on the same frozen prompt set.

Result

Citation share from 3% to around 60% within one quarter. Brand mentions in AI answers up more than tenfold. When the AI grounds its answer in your own page, it almost always speaks positively about your brand.

Status: presented in the neuroflash × Rankscale webinar (21 May 2026, with Patrick Schmid of Rankscale). Company anonymised.

Read the webinar recap

Case D

wireworks.wiki

A fan wiki for an indie game that our CIO set up as a second test bed: new domain, no history, live since 11 August 2026.

0% 31.3%
ChatGPT mention rate on 150 fixed queries, 11 August to 6 September 2026, measured with Rankscale.

Starting point

A new domain with no backlinks and no history. Before launch we froze 150 fixed player queries in Rankscale: 0 citations, 0% mention rate. ChatGPT answered from the game’s Steam page and the Steam community.

What we did

Nine guides at launch, sitemap submitted, comments under eight YouTube videos about the game. After ten days, the titles and meta descriptions of pages on the edge of page one were rewritten. Measured every two weeks on the same query set, each run started manually.

Result

ChatGPT mention rate from 0% to 31.3% in 26 days, meaning 47 out of 150 answers. Citations from 0 to 85 across six URLs, average position 1.1. On 6 September the wiki was cited more often on its query set than the game’s own Steam page: 341 mentions to 338.

Status: a new wiki, measured honestly, not a controlled sample. Google AI Overviews were not queried in the third run.

Visit the wiki
Book a free intro call

Four ways to get started

The first three options contain the same package; what differs is how much of the setup our team does with you. With Managed, the loop runs on our side.

What it costs the usual way

Swipe sideways to compare.

Task Freelancer or agency You, with your agent Managed by neuroflash Output
Real customer questions Keyword tools and guessed prompts Digital Twins built on over 1 million survey profiles We generate them every week from your twins Queries the way people actually type them
Website build €5,000 to €15,000, 4 to 8 weeks Included in the tooling, days instead of weeks Stays with you or your team A site you keep extending
Website changes and hosting €50 to €150 per month €0 to €20 per month Stays with you; we deliver copy and structure Unlimited same-day changes
Short clips from videos €150 to €400 per clip Minutes per clip Not part of Managed 10 to 20 clips per video instead of 2 to 3
LinkedIn ghostwriting €500 to €1,500 per month €50 to €100 per month in tools Not part of Managed Posts every week, in your voice
Total running costs €550 to €1,650 per month €200 to €400 per month €1,999 per month Easily 10× the output

Typical ranges at European market rates, not quotes. A year of tooling costs €2,400 to €4,800. For most businesses, one new client covers that. Managed costs €1,999 per month with a 3-month minimum term.

Course

€499

Learn at your own pace. The complete agent framework as a self-study course: setups, prompts and workflows for every result on this page, from website to book draft.

Buy now · €499 Talk it through first

Cohort

€1,999

A small group, live over several weeks with our team. Feedback on your actual website, your clips and your posts.

Guarantee: within 3 months of completion, ChatGPT names you when asked about your name and your topic. If it doesn’t, we keep working with you.

Book now · €1,999 Talk it through first

Most intensive

Done-with-you

€5,000

We set up the agent framework with you, 1:1. Full materials, up to four one-to-one sessions, and a direct line to us for three months.

Guarantee: within 3 months of completion, ChatGPT names you when asked about your name and your topic. If it doesn’t, we keep working with you.

Book now · €5,000 Talk it through first

We run it

Managed

€1,999 / month

3-month minimum, then month to month

You measure every week, we deliver every week.

  • Weekly scan of your frozen query set (ChatGPT, Gemini, Perplexity)
  • Twin questions: new real customer questions from your Digital Twins
  • 4 answer-first pieces per month, written and delivered for your approval
  • Monthly gap briefing with a visibility report
  • Who touches your website: you or your team; we deliver copy and structure

Scoreboard: your query set, measured every week.

Book an intro call

* The guarantee applies to Cohort and Done-with-you. For Managed, your frozen query set is the scoreboard, measured weekly, with a 3-month minimum term. Prices in EUR plus VAT. Ongoing tool costs typically run at €200 to €400 per month. In most cases, one new client covers the year.

Frequently asked questions

What exactly is in the package?

Access to neuroflash Digital Twins via MCP, together with skill files and process documents for your coding agent: audience definition, query generation, visibility audit, GEO article writer, website editor with rollback, LinkedIn engine, copywriting system, image generation and video repurposing, plus a 7-day setup programme. Your agent reads them and runs the loop.

Why Digital Twins? Why not just ask ChatGPT for the questions?

A standard LLM generates queries of 15 to 25 words in perfect grammar. Real users type a median of 11 words; 46% start in lower case and 12% are bare keyword fragments. Digital Twins are calibrated to this distribution and grounded in over 1 million real survey profiles. Only content that matches real questions gets cited.

What makes ChatGPT cite a page?

Semantic fit between the real question and a passage on your page. Each standard deviation raises the odds of being cited 3.27-fold (95% interval 2.68 to 4.01). Putting the answer in the first 60 words brings 2.02×, a title that uses the words of the question 1.96×. High list and table density works against you: 0.89×.

How long until ChatGPT recommends you?

The loop runs weekly, with a measurement after seven days. On jonathanmall.com the mention rate rose from 3.7% to 26% in nine weeks, meaning 26 out of 100 answers in the top-100 query set. That is a single website, measured honestly, not a controlled sample.

Do backlinks help?

Not measurably in our study. Backlinks showed no independent effect (p = 0.42 and p = 0.15). They remain relevant for classic search, but our study found no evidence that they raise the odds of being cited in AI answers.

Does FAQ schema make a difference?

On its own, it showed no reliable effect (p = 0.27). It is still worth having as cheap hygiene, but it won’t work as a lever.

How many queries does a reliable measurement need?

Hundreds. Our study ran on 1,362 queries, our ongoing pulse on over 1,300. Smaller sets fluctuate so much that you mistake chance for progress.

What is the difference between Done-with-you and Managed?

Done-with-you is a one-off setup: we build the framework with you, and from then on your agent runs the loop itself. With Managed, the loop runs on our side. Every week we scan your frozen query set in ChatGPT, Gemini and Perplexity and generate the twin questions, and each month we deliver four answer-first pieces and a gap briefing. The minimum term is three months.

About neuroflash

neuroflash is the AI marketing suite from Hamburg. Our Digital Twins are AI-powered representations of real people, built on up to 255 data points per person from over 1 million real surveys since 2017, with 85 to 98% accuracy. Available via app, API and MCP.

Teams at REWE Group, Migros, Swisscom, StepStone and fischerAppelt work with our twins. Around 20 million real user queries on our platform form the calibration base for the queries generated in the AI visibility loop.

Digital Twins at neuroflash The full study, including methodology

A study by our CIO Dr. Jonathan Mall, published on his own website.

Get found, starting with the first call.

Book a free intro call

Sources and methodology

Our pre-registered quasi-experiment with 26 pre-declared variables: 1,362 queries, 2 engines, 3 snapshots from 2 to 16 July 2026, 52,821 citations, 847 retrieved pages, AUC 0.874.

The website figures come from Google Search Console, PageSpeed Insights and the visibility pulse of 17 August 2026. We checked the realism of the queries against around 20 million real user queries on our platform.

The Digital Twin figures (over 1 million real survey profiles since 2017, up to 255 data points per person and 85 to 98% accuracy) come from neuroflash.com/en/digital-twins.

The findings are correlational, not causal. Only the persistence model with n = 7,701 comes close to a causal statement.