Service ·
Actions and impact: what we change, and whether it works
In short
anewera turns the AI's answers into concrete actions: change your own pages, contact sites that cite your competitors, and write missing content. Once an action is done, we compare the same questions on the same AI services before and after. An effect only counts when the margin of error of the difference no longer includes zero.
Service
How we get you into the answers
So that ChatGPT, Claude and Gemini recommend your company when your customers ask: four steps that repeat every month. At the end there is proof of what worked.
Analyse
Every day we ask the questions your customers ask and read every answer: are you named, recommended, cited as a source? If not: who does AI name instead, and which pages does it read?
ChatGPT Who can compare supplementary health insurance plans for me in English?
My top recommendations around Zürich
1. Expat Savvy (Best starting point)
Zürich · Advice in English
Offers a free 45-minute consultation, comparisons across insurers, and a written recommendation.
Recommended, named firstCited: expat-savvy.ch
Change
We change exactly the pages and sources AI builds its answer from: technology, content, listings and mentions. Every change gets a date.
Changes - Region pages: the first sentence names ReloFinder and the number of agencies, in English, German and French14 Oct
- Agency profile Welcome Service: rated 4.7 from 41 reviews14 Oct
- Corporate clients: the answer for HR teams at the topIn progress
- Cost guide: the short answer with the price range firstNext
Measure again
The same questions keep running, day after day. Every month the report shows where you stand against your competitors and how AI names you.
Report for Expat Savvy Named in the answers
How AI names Expat Savvy
- Recommended43%
- As a source42%
- Listed12%
- With a caveat3%
Cited most
- expat-savvy.ch45%
- bag.admin.ch20%
- en.comparis.ch19%
Prove
Before against after, with the margin of error. An effect counts only when it is larger than chance. What does not work, we drop.
Before and after Changed on 14 Oct: region pages
ReloFinder named, in questions about agencies in a city
Before8 to 14 October 202647% ±4575 answersAfter8 to 14 November 2026Measuring again
Every engagement starts with a first conversation. We then agree the scope together, to fit your market and your goals. Book a first meeting
- How we work in detail
- We show how we measure in the open, with a service run by Loaded AG, the company anewera belongs to. Read the case study
- Grow in Public: How we make anewera visible, from zero
Four kinds of action
Every action comes from the measured answers, not from a checklist:
| Action | When it pays |
|---|---|
| Change a page | one of your pages is read for a question but not cited, or the AI states wrong facts there |
| Contact a site | another site is cited for your questions and names your competitors, but not you |
| Ask a partner | a partner or association is cited and could mention you |
| Write content | you are rarely named for a question, and no good page answers it |
The view also shows why the AI names your competitors: with the sentences it uses to justify their recommendation. Sites where a request achieves little sit on a separate list to review.
For page changes and new content, every action comes with a ready brief: what the AI cites and says, the facts that must be right, the rules that keep the page from losing in Google, and how the effect is measured. Your team or your agency can work from it directly. Requests to other sites we draft, but never send unasked.
How we measure the effect
Every action has a status: open, contacted, done or dismissed. Once it is done, we compare:
- Before and after: the share of answers naming you, up to 28 days before and up to 28 days after the date.
- Pairs only: the same question on the same AI service, each pair counting the same. New or removed questions cannot skew the comparison.
- The right questions: when the action concerns a particular page, the questions whose answers read or cited that page count. Otherwise all questions, and the page says so.
- With a margin of error: from a bootstrap over the questions.
The result reads better, worse, not clear yet, too few pairs (under five) or not measured since. Next to it we show whether the changed page is read and cited more often since.
Why with a margin of error
AI answers vary from day to day, even when nobody changes anything. A rise over a few days can be chance. With us an effect only counts when the margin of error of the difference no longer includes zero. Until then it says “not clear yet”, even when the number already looks better. What does not work, we drop.
An example from our work
Above you see the four steps of our work on real data: a measured answer, the actions on relofinder.ch with their dates, the report and the running before-and-after comparison. relofinder.ch, like anewera, is a website of Loaded AG.
What it costs
The actions and the measurement of their effect are part of Analytics, from CHF 290 a month. With Managed our team carries out the actions for you, from CHF 980 a month. As a one-off there is the GEO audit for CHF 1,900. All prices are on the pricing page.
Frequently asked questions
How long does it take for GEO measures to work in ChatGPT and Gemini?
For answers from web search, often days to weeks: as long as it takes until the changed page is read again. What a model learned in training only changes with a new model version. So we compare up to 28 days before and after an action and say “not clear yet” as long as the data allow no conclusion.
Can an agency guarantee that ChatGPT recommends my company?
No. Nobody controls what ChatGPT answers. An agency can promise what it does and how it measures. We show every action with its date and its effect with a margin of error, even when it achieved nothing.
How does anewera measure whether an action works?
We compare the share of answers naming you up to 28 days before and up to 28 days after the action. Only pairs are compared: the same question on the same AI service, each pair counting the same. When the action concerns a particular page, the questions whose answers read or cited that page count. The margin of error comes from a bootstrap over the questions.
Which actions does anewera suggest?
Four kinds: change one of your own pages, contact another site that is cited for your questions but does not name you, ask partners for a mention, and write content for questions where you are rarely named. For page changes and new content there is a ready brief for your team or your agency.