Asks ChatGPT, Perplexity, Gemini and Grok your customers’ questions and shows who they recommend: you or your competitors. Finds errors about you and the sources behind the answers, prepares fixes, and repeats the measurement.

A customer asks ChatGPT: “Which CRM should I choose for a small store?” The AI names three of your competitors. You're not in the answer, and you don't even know it. Vera asks AI models your customers' questions, shows who they recommend, and prepares fixes. This work is called GEO: optimization for AI-generated answers.
In an internal ASCN test, we pasted ChatGPT's answer to “Which no-code platform should I choose for AI agents?” into Vera's chat. Vera broke it down right away:
The AI named Make, n8n and Zapier. All three are on your competitor list. ASCN isn't in the answer.
To record this measurement correctly, please specify: the date the question was asked, the region, and the links under the answer. AI answers change, and without this data the measurement can't be repeated and compared.
Yes, even for the company that builds Vera, the AI recommends competitors. That's exactly why one question in ChatGPT settles nothing: you need regular measurement, an understanding of where AI models get their answers, and concrete fixes.
You published a new page and updated your directory profile, but whether it helped is unclear. You asked ChatGPT once and it named you, and the next day it didn't. A couple of weeks later, Vera asks the same questions to the same AI models and compares:
“Three weeks ago, Perplexity named you in 3 answers out of 30; now it's 9. ChatGPT is unchanged: 1 out of 30. The growth coincided with publishing the comparison page, but a third-party review mentioning you came out in the same weeks, so the result can't be credited to a single change.”
AI models don't invent recommendations: they rely on specific sources like Wikipedia, reviews, directories, YouTube and review sites. Vera finds which sites the answers link to, where your competitors are present and you're not, and prepares the content: an answer page, FAQ, comparison, directory profile.
“In 14 of 30 answers, the AI models link to the same software directory. Three competitors have profiles there, you don't. Here's a ready-made profile: description, pricing, answers to common questions. You publish it.”
For an agency, telling a client “we worked on your AI visibility” isn't enough. You need proof. Vera saves every AI answer with the date and a quote, and shows for the client and 3–5 of their competitors: recommended, mentioned, or not named at all. A repeat measurement a couple of weeks later gives you a before/after comparison you can put in a report, without crediting success to a single change.
Old prices, wrong facts, a product you discontinued two years ago. The customer believes the AI's answer and leaves without ever opening your website. Vera finds errors about you in AI answers and shows where they come from.
“Perplexity lists your plan at $49, though it's been $29 since March. The source is a 2024 review on a third-party site. I suggest updating your pricing page with an ‘as of’ date and writing to the review's author. A draft email is ready.”
Years of SEO investment, first page of search results. But customers increasingly ask an AI instead of a search engine. SEO is about your position in search; Vera checks who the AI names in its answer. For example, here's a measurement across 30 customer questions:
| AI model | You were named | Main competitor was named |
|---|---|---|
| ChatGPT | 2 of 30 | 14 of 30 |
| Perplexity | 3 of 30 | 12 of 30 |
| Gemini | 1 of 30 | 11 of 30 |
| Grok | 1 of 30 | 9 of 30 |
The measurement comes with a list of questions where the competitor is named and you aren't, and what to add to change that.
Send your website and 3–5 competitors → Vera drafts 20–30 questions your customers ask, and you approve them → using an OpenRouter key, Vera asks these questions to ChatGPT, Perplexity, Gemini and Grok and saves every answer with the date and a quote → shows who's recommended, who's mentioned, who isn't named, and which sources the answers rely on → finds errors about you and prepares fixes as drafts → you publish → a couple of weeks later, Vera repeats the measurement and shows what changed.
| Asking ChatGPT yourself | Vera | |
|---|---|---|
| How many questions | One or two, whatever comes to mind | 20–30 questions your customers actually ask |
| How many AI models | One | Four: ChatGPT, Perplexity, Gemini and Grok |
| Reliability | Tomorrow the answer will be different, and you won't know | Every answer saved with the date and a quote, so the measurement can be repeated and compared |
| Competitors | You only see what happened to appear in one answer | You and 3–5 competitors: recommended, mentioned, not named |
| Why | Unclear | The sources the answers rely on, and where you're missing from them |
| What to do | Generic advice like “improve your content” | Ready-made content: answer page, FAQ, comparison, directory profile |
Asking ChatGPT is a single snapshot. Vera shows how the picture changes and what to do to change it in your favor.
Mattress manufacturer. Everything worked right away. They connected an OpenRouter key and approved 25 questions like “which mattress should I choose for back pain.” Vera found that ChatGPT and Gemini were naming a discontinued model and a two-year-old price. The source was an old review on an aggregator site. The brand updated its product pages, added an FAQ and wrote to the review's author. A month later, the error was gone from Perplexity and Gemini answers, and remained in 2 of 25 ChatGPT answers.
Digital marketing agency. This one took some work. At first they worked without an OpenRouter key: managers asked the AI models questions themselves and pasted the answers into the chat. It was slow, and Vera kept asking for the date and region, without which measurements can't be compared. Two weeks later they connected a key and agreed on a request limit. Now the agency runs a monthly measurement for four clients and shows them a before/after comparison with quotes.
CRM for beauty salons. It worked partly. Vera showed that competitors were being named because of two comparison articles and one software directory. The company published a comparison page and a directory profile. Three weeks later, Perplexity named it in 9 answers out of 30 instead of 3, while ChatGPT still named it in only one. A Wikipedia article wasn't possible: the company isn't well-known enough yet, and Vera said honestly that this route is closed for now.
For measurements, you need an OpenRouter key. It's a service where one key gives access to different AI models: Vera uses it to ask ChatGPT, Perplexity, Gemini and Grok. Requests are billed at OpenRouter's rates, separately from ASCN. Before each measurement, Vera shows how many requests there will be and stays within the agreed limit. Without a key, Vera analyzes answers you paste into the chat.
Vera can also: check whether you're on Wikipedia, YouTube, review sites and directories that AI answers link to, and prepare a profile where you're missing.
GEO (Generative Engine Optimization) is the work of making sure AI models like ChatGPT and Perplexity know about you, name you in their answers, and say true things about you. Vera measures who AI models recommend to your customers, finds what the answers rely on, and prepares fixes.
No. SEO is about your position in search. Vera checks who the AI names in its answer.
Nobody can guarantee it. What helps is what AI models rely on: clear pages with answers, up-to-date facts about you, directories and reviews. Vera finds what's missing in your case and prepares the content.
No. Content for pages, directories and emails to review authors arrive as drafts. You publish and send.
No. Vera works with your data only inside your ASCN account, to carry out your tasks. To get answers, customer questions are sent to AI models through your OpenRouter key. You can revoke access at any time in the integrations section. Learn more about how data is stored in the ASCN privacy policy.
Requests are billed at OpenRouter's rates, separately from ASCN. Before each measurement, Vera shows how many requests there will be and stays within the limit you've agreed on.
Yes. Ask the AI models questions yourself and paste the answers into the chat, and Vera will analyze them. But it takes longer, and for accurate comparison you need to include the date, region and links under each answer.
Not yet. Measurements run through ChatGPT, Perplexity, Gemini and Grok.
Vera asks the same questions after your changes and shows what changed. Success isn't credited to a single change: if something else happened in the same weeks, Vera will tell you.
It depends on your niche, competitors and how many fixes you publish. Typically, after three months you have a dated history of measurements, an understanding of which sources AI models rely on in your niche, corrected errors about you, and profiles where you were missing. Nobody can guarantee you'll start being named, but you'll see the changes in the measurements.
Message ASCN support: we'll help you connect an OpenRouter key and work through your first measurement.
Most often she's missing data: a competitor list, approved questions, or the date and region for pasted answers. Ask Vera herself: “What do you need for this measurement?” — she'll tell you. If she's still confused, start a new session: this clears the context, and Vera will start from a clean slate.
