Mention RateShare of VoiceCitationsSentimentSource MapCrawler Access
why audit first
You can't fix what you haven't measured
Chat assistants don't appear in your analytics as a channel with positions and clicks. A buyer asks a question, reads a shortlist and may never open your site. Most companies have no idea whether they are on that list, or what the model says about them when they are.
The audit gives you that baseline. We collect the questions your buyers really ask, run them across the major assistants and record who gets named, who gets a link and which websites shape each reply. Then we check your own site and reputation to explain the result, and turn the findings into a plan.
MENTION
Named in the reply
The assistant names your company. That builds awareness even without a visit.
CITATION
Linked as a source
The assistant links to one of your pages. That brings traffic and signals trust. We count the two separately, because each needs a different fix.
LEARNED
Known from memory
The model knows you from its training data and recommends you even with web search switched off. This reflects long-term reputation.
FOUND
Found by live search
You show up only when the assistant searches the web. This depends on indexing and fresh content. Where a model lets us separate the two modes, as ChatGPT, Gemini and Claude do, we test both.
51%
of B2B software buyers now start research in an AI chatbot more often than in Google, up from 29% a year earlier.
G2, The Answer Economy, 2026
69%
picked a different vendor than planned after a chatbot conversation. A third bought from a company they hadn't known.
G2, 2026 · 1,076 buyers
64%
of those buyers often run into inaccurate recommendations. Errors about your brand cost deals you never see.
G2, 2026
57.5%
of US consumers who use AI have dropped a purchase because of what a chatbot told them.
Semrush / Exploding Topics, Jul 2026 · among AI users in a survey of 2,338 US adults
Global and US surveys. Your own numbers come from the audit.
what we check
Eight questions the audit answers
Are you named on buyer questions?
Your mention rate on a prompt set built from real demand in your category.
Which assistants know you?
Results for each model, from memory and with live lookup wherever the two can be separated.
Who wins instead of you?
Share of voice against direct competitors on identical prompts.
What do they say about you?
Tone, recurring themes and how accurately your products, prices and locations are described.
Which sources shape the reply?
A map of cited domains and pages, grouped into your site, media, review platforms, forums and directories.
Which sources mislead?
Pages with outdated or wrong facts about you, each with a suggested action.
Can AI crawlers read your site?
robots.txt rules for GPTBot, OAI-SearchBot, Claude-SearchBot, Google-Extended and others, plus server-side rendering and schema markup. llms.txt is noted too, as a proposed format rather than a standard.
What should you fix first?
Tasks ranked by impact and effort, ready for your team or ours.
how it compares
SEO audit, AI checker or GEO audit?
Three tools that sound alike and answer different questions. Most companies need the first and the third; the second is a quick signal between them.
SEO AUDIT
ONLINE AI CHECKER
GEO AUDIT BY ZUM PUNKT
Main question
How does the site rank in Google?
Is the brand mentioned at all?
What do assistants say about you, why, and what to change
Where it looks
Your site and its search positions
A few AI answers
AI answers, the sources behind them and your site
Prompts
Keywords, not prompts
Generic or self-typed
Built from real demand and the customer journey
Runs and modes
—
Usually one run
Repeated runs, memory and live search, by market
Competitors
Ranking overlap
Rarely
AI share of voice on identical prompts
Off-site sources
Backlinks
—
Recurring cited domains, checked for accuracy
Output
Technical fixes
A score
Reasons, a ranked plan and a team to run it
who it's for
Who the audit is for
Companies whose customers compare options before getting in touch: complex services, expensive purchases and crowded markets.
See how chatbots position you against rivals long before it shows up in sales numbers.
Marketing & SEO
Get the prompts, gaps and pages to work on, plus a baseline to report progress against.
PR & Communications
Learn which media and rating sites the models trust in your niche, and where the story about you is wrong.
Product & Sales
Hear how assistants describe your offer, pricing and alternatives to buyers.
what you get
What's inside the report
Prompt map30–50 buyer questions weighted by demand and grouped by intent and product line.
Visibility snapshotMention rate, citation rate and share of voice for each assistant, compared with up to five competitors.
Brand framingSentiment, recurring themes, awareness level from “unknown” to “category leader”, and factual errors.
Source mapThe domains and pages behind the replies, each with an accuracy score and a next step: update, request a mention, earn a link or keep monitoring.
Site readiness checkCrawler access, rendering, structured data, entity clarity and answer-first copy on key URLs.
90-day planRanked tasks across content, technical SEO, PR and reviews, with owners and expected effect.
Walkthrough callWe present the findings to your team and answer every question.
ONE FINDING FROM A REPORT
PROMPT“Best CRM for a 200-person B2B team?”
FINDING
Competitor named in 5 of 6 runs, Lumen CRM in 1 of 6.
SOURCE
techradar.com review from 2024 with outdated pricing.
ACTION
Request an update · Add a pricing table to /pricing
Sample data. Your report uses your brand, competitors and markets.
how it works
How we run the audit
1BriefYour products, markets, languages and main competitors. We may ask for read-only webmaster, analytics and server log access.
2Prompt mapQuestions come from search demand, sales calls and support requests, not from a generic template.
3Multi-run testingEvery prompt runs several times per assistant, market and language, with web search on and off where possible.
4Source and site analysisWe trace each citation, check the facts and review your own pages and crawler settings.
5Report and planFindings, ranked actions and a call to walk your team through them.
6Re-auditSame prompts, same method, so the numbers stay comparable over time.
A free check takes one business day. A full audit takes 2–3 weeks, depending on markets and languages.
The method in detail
Customer journey first. A buyer talks to a chatbot differently at each stage, and a brand can win one stage and vanish at the next. So the prompt map follows your customer journey map (CJM), and the report shows the stage where you drop out.
Problem“Why do our sales reps keep losing deals in spreadsheets?” Do you appear as the expert who explains it?
Category“What types of CRM fit field sales teams?” Is your type of product named?
Shortlist“Best CRM for a mid-size B2B company” Are you on the list?
Comparison“Lumen CRM vs Brand B for a 50-person team” Who wins, and on which arguments?
Validation“Is Lumen CRM reliable? Any complaints?” What reviews and facts come up?
Purchase“Lumen CRM pricing, contract terms, how to migrate” Are the details correct?
Query fan-out. Before answering, an assistant splits one prompt into several background searches. Where the engine exposes them or lets us reconstruct them from its citations, we log these sub-queries and check which page answers each one in Google, Bing and Brave. A missing page is a concrete task.
ONE PROMPT
“Which CRM fits a B2B sales team moving off spreadsheets?”
BACKGROUND SEARCHES · YOUR COVERAGE
CRM for mid-size B2B companies 2026✓ /solutions/b2b
CRM pricing per user comparison✗ no page
CRM with ERP integration reviews✓ G2 profile
Lumen CRM vs competitors✗ outdated review
Sample data. Each ✗ becomes a task in the plan: a page to create, a listing to update or a mention to earn.
Clean conditions. Fresh sessions without chat memory, the target country and language, and the same wording every time, so a change in the numbers means a change in the market, not in the setup.
two levels
Start free, go deeper when you're ready
FREE VISIBILITY CHECK
FULL AI VISIBILITY AUDIT
Prompts
10–15 key questions
The full prompt map, weighted by demand
Assistants
ChatGPT, Gemini, Google AI Overviews
All of them; memory and live search where the assistant allows it
Send us your site and two competitors. We'll run the free check and show you the first gaps.
20+
years in digital
500+
projects delivered
50+
specialists
2005
enterprise clients since
An online checker gives you a score. We give you the reasons behind it and a team that can act on them, with search, reputation and content under one roof.
clients about us
What clients say
We receive reports on completed work and recommendations for improving our site's pages on time.
Valery IlyinskyCEO and co-founder, Genotek
We look closely at the reports the specialists prepare, and we are happy with the result.
Vera MalakhovaDevelopment Director, Milano aesthetic medicine clinic
We get exhaustive answers to our questions. It is interesting to dive into the deeper mechanics of promotion.
Alexey NovikovDirector, PARADA medical clinic
Translated from reviews published on zumpunkt.ru.
50+
50+ specialists. Zero generalists.
Scripts collect the runs; GEO analysts write the prompts and read the replies, so the conclusions come from people, not from a score. SEO tech leads run the site checks, and our reputation team handles source analysis.
An SEO audit looks at how your site ranks. This one looks at what chatbots say about your company, which outside sources they rely on and whether they can read your pages. The comparison table above shows the difference point by point.
2.Which assistants do you test?
All the assistants listed at the top of this page. DeepSeek and other models can be added on request.
3.Can you guarantee we'll appear in AI answers?
No, and nobody honestly can: the models decide, and they change often. What we are responsible for is the accuracy of the measurements and the quality of the recommendations. The method is documented, the prompts stay fixed and you get the raw replies, so every number can be checked.
4.Why run each prompt several times?
The same question can get different replies within minutes. One screenshot may be luck. Repeated runs reveal a pattern you can act on.
5.Can we check it ourselves?
Yes, and it's a good start: ask a dozen buyer questions in ChatGPT and Gemini and note who gets named. It just won't go far. Your own chat history and location colour the replies, single runs are noisy, and you won't see the background searches the engine keeps hidden, the full source picture or your share against rivals. The audit adds those layers.
6.Isn't a free online checker enough?
It gives a quick signal. It won't explain why a rival wins, which sources mislead or what to fix first. That is what the full audit covers.
7.What if assistants don't mention us at all?
That's common, especially for younger brands and niche B2B suppliers, and it is a result in itself. The audit shows the reason: pages that aren't indexed, crawlers that are blocked, an unclear entity, or no mentions on the platforms assistants rely on. The plan then starts from those foundations.
8.Do you check llms.txt?
We note whether you have one, but we don't score it. llms.txt is a proposed format, not an accepted standard, and none of the major assistants has confirmed using it for answers. Crawler access, rendering and sources matter far more.
9.Can you work with our current SEO agency?
Yes. The audit is vendor-neutral: tasks are written so that your team, your SEO contractor or we can carry them out. Many clients hand the technical part to their agency and keep us for sources and reputation.
10.What access do you need?
For the free check, none: the website and a couple of rivals are enough. For the full audit, read-only access to your Google and Bing webmaster accounts and your analytics, plus a server log export to see AI crawler visits. Everything is optional, but it makes the findings sharper.
11.How long does it take?
A day for the free check. The full audit needs two to three weeks; every extra market or language adds time.
12.Can you cover several countries or languages?
Yes. Replies differ by market, so each one gets its own prompt set and its own results.
13.How often should we repeat the audit?
Quarterly for most companies, and 2–4 weeks after major content or PR changes. We keep the same prompts so the numbers can be compared.
14.Do we get the raw data?
Yes. You receive the prompts, the full replies, the background searches we logged and the source list, not just a summary.
15.What comes next?
Your team can follow the plan on its own, or Zum Punkt can take it over. The audit price does not depend on what you decide.
16.How much does a full audit cost?
The price follows the scope: prompts, markets and languages. After the brief we send a fixed quote.
full audit
Start with a brief.
Tell us about your company, markets and competitors. We'll suggest the right scope and reply within one business day.
1Brief · 2 minutes
2Prompt map · your buyers' real questions
3Testing · six assistants, several runs
4Report · findings and a ranked plan
Thank you. Your request was accepted. We’ll be in touch within one business day.
