Who benefits from working with their own agent
This workflow suits businesses that already have a website and access to a compatible agent, such as Codex, Claude Code or another client supporting remote MCP. The agent can investigate pages and make changes, while the owner needs a clear plan and a history of the outcomes. For multiple websites, each business’s tasks and results are kept separate.
In AURA GEO, you can choose ChatGPT, Gemini, Perplexity and other directions in the workspace. This selection defines the focus of the work. To obtain a real answer from a particular service, the agent uses an interface and session available to it. Simply recording a question in AURA does not launch a check in a third-party chat.
Getting started: a website link, subject area and customer questions
The owner registers in AURA and connects the MCP server at https://aurageo.io/mcp. A website can be added in the workspace or delegated to the agent by providing a link. After authorization, the agent gets access to selected projects and can continue setup without making the owner search for every section manually. Detailed connection instructions are in the AURA MCP documentation.
First, describe the business accurately: its public name, services, audience, country and language. Then choose questions a potential customer asks before learning about the company. For example: “Which tools help check website recommendations in ChatGPT?” Questions containing your own brand name are useful for a separate check, but do not replace this measurement. The owner should confirm the subject area and question list: similar names and abbreviations may refer to different markets.
An audit becomes specific changes
The agent retrieves website preparation tasks and checks public pages: server response, accessible HTML, crawl rules, canonical, indexing restrictions, sitemap and markup. The audit aims to find a specific obstacle and suggest a fix. An open page with clear content gives search engines material to read; a useful answer relevant to the customer gives them a reason to use that material.
For ChatGPT, it is important to distinguish the bots. OpenAI describes OAI-SearchBot as a search crawler, GPTBot as a bot for potential data use in training, and ChatGPT-User as requests resulting from user actions. Review configuration together with the CDN and server security. Using a bot’s name in a technical request can reveal some blocks, but does not verify a genuine OpenAI visit.
The Pupupul review in the AURA library demonstrates this approach: it checks page facts, content accessibility and inconsistencies in the product description. It is an audit example, not a claim of AI recommendations achieved.
How the agent gets access to files and publication
MCP connects the agent to an AURA project. To change an existing website, the owner must separately provide an authorized way to work: a repository, files, CMS or deployment access. The agent reviews the project structure, makes appropriate changes and checks the published result. If the required access is missing, it shows the next step and necessary link rather than marking setup complete.
For managed content, AURA supports an /answers/ section and a subdomain option after a connection is installed on the owner’s website. The agent prepares a page using real facts and sources. Publication depends on domain verification and granted permissions: keep drafts only, approve changes or allow autopilot within defined limits. The owner controls access and can revoke it. Signing in to Google or Bing, granting permissions and paying external services remain owner actions.
What happens after publication
Finished content should address a distinct visitor need. Begin with a direct answer, followed by conditions, facts, examples and sources. Similar question formulations often need just one good article. Producing many similar pages and repeating keywords does not replace substantive work. For AURA GEO, an article might explain implementation using your own agent, supported by an actual review with completed actions and check results.
After publication, AURA lets you check the live HTML and submit changed URLs through IndexNow, and sitemaps through available search engine connections. IndexNow documentation defines this as a change notification, not the purchase of a position or confirmation of a page’s inclusion. The workspace therefore records publication, submission, search engine indexing information and observed citations separately. Page content and search status can be improved and checked step by step.
AI checks: every answer can be opened
The workspace prepares a run with fixed questions, selected AI services, the requested market and language. The agent retrieves the task, submits the original question in an available service and saves the full answer, source URLs, time and mode or model actually indicated. If the interface does not allow geographic settings, note this in the check conditions. The history shows which tasks received answers and which need sign-in, permission or restored access.
AURA distinguishes a brand name in text, a link to its domain and explicit recommendation context. An answer mentioning a company does not necessarily suggest choosing it. Question wording is saved for comparison; a post-edit check is meaningful under matching conditions. Variants with a different model or retrieval method should not be presented as the isolated effect of a content change. Periodic runs are performed by an operating external agent within its available sessions and limits.
Our baseline: one question, three services
On 18 October 2026, we checked a neutral buyer question for AURA GEO: “Which services help promote a website in answers from ChatGPT, Gemini and Perplexity? Recommend options and link to their official websites.” Answers were obtained in the ChatGPT, Gemini and Perplexity web interfaces; full texts and sources were saved in the workspace.
These three answers did not mention AURA or link to aurageo.io. This is a baseline result for one question under specific conditions, not a conclusion about every possible query or the website’s absence from a search index. The next work focuses on the sources found and the usefulness of our own pages: clarify the product workflow, show a verified case and then repeat a comparable measurement. Growth or successful citation can be claimed after obtaining the corresponding answers.
How to see business value
A citation shows that a saved answer contains a website link. To see visitor arrivals, install a tracker on public pages; AURA managed pages include it automatically. For the rest of the website, the agent installs the supplied code snippet through authorized access and verifies that events actually arrive. Attribution uses available referral data, so some visits may remain unrecognized.
The owner can access answer history, sources, publications and agent actions. These help explain what changed and which task to do next. Leads require a separately configured event, while bot visits require server observations. Link counts in answers do not automatically become customer counts. Ongoing work involves checking new answers, resolving confirmed problems, updating outdated facts and assessing real visits.
What AURA GEO costs and where to start
The pricing page presents three workspace plans: Start — €0 and one website; Growth — €29.99 per month and up to three websites; Business — €99.99 per month and up to five websites. Paid plans include the AI directions without a separate price per model. Each website retains its own work history.
Early access is currently in effect: tools are available to explore, plan selection is saved without charges, and payment acceptance has not launched yet. The described workflow does not require paid model APIs: it uses available sessions in your own agent and free search engine tools. The agent’s and AI services’ own limits still apply.
Start with one website and one clear workflow: confirm the business description, connect the agent, check baseline answers and make the first justified improvement. This builds a workspace history of completed work and verifiable results to support your next decision.