A small category does: an agent installed on the site itself, not a dashboard next to it. It proposes the exact change, shows it in full, and applies it on your click, the same way on a second site and a third. Most of the category still stops at a report or a suggestion someone has to carry out by hand.
This is the gap buyers keep describing without a name for it: a tool told them what was wrong, and nothing changed afterward. The full inventory of what an agent like the one inside Reflexa actually does, and does not, on your site is below.
It means the tool itself changes something on the live page, not that it describes the change for a person to make. Most of the category stops one step earlier than that. A tracker asks ChatGPT, Claude and Perplexity a set of buyer questions and reports how often your brand gets named; nothing on your site moves because of it. An SEO suite's newer AI module reads your pages and lists what it thinks is wrong, as a suggestion in a dashboard, the same shape as a crawl-error report. Both are useful for seeing the problem. Neither one is the fix.
| Tier | What it does | Who still has to act |
|---|---|---|
| Tracker | Counts mentions, citations and recommendations across engines | You, on the site, with no help from the tool |
| SEO suite's AI module | Flags a page or a schema gap as a suggestion | You, or a developer, still editing the CMS |
| Site agent | Proposes the exact change and applies it on your click | Nobody, once you approve it |
Why it matters: the jobs a tracker or a suggestion leaves behind are usually too small to brief to a developer and too repetitive to do by hand across every page, so they stay undone for months even after the report names them correctly.
Three levels of trust, each shown in full before anything
happens. Four switches the agent flips itself: AI crawler
access in robots.txt, Organization structured data in
the page head, the sitemap, and an /llms.txt built
from your own page titles. Page text it writes for your approval
first: a title, a search description, an answer-first opening, and
an FAQ block with its schema. And since October 2026, a general
change hand for almost anything else on the page, a stylesheet, a
snippet, a passage, a new page, or a response header, each
proposed in full and applied only on your click.
On a coded site the agent's hand is one edge file in front of it, Cloudflare, Vercel, Next.js, Netlify or Node. On a site whose code lives in a repository, a GitHub App install or a GitLab project token turns every approved change into a commit on your default branch, undo as another commit. See the full inventory of what the agent changes and never does for the complete list.
Do this: most tools stop reporting at "applied". Reflexa also records a baseline the moment a change lands and reads it again fourteen days later, returning a verdict in plain words: moved, flat, or worse. Read how a change is checked to see if it actually worked.
It is not a browser agent clicking through your site, and it is not a guarantee that AI will name you afterward. Reflexa writes through whichever surface it is installed on, the plugin's own API on WordPress, the edge file, or a commit on a connected repository, never by simulating a person clicking buttons in a browser. It is also coded to never do eight things, whatever is asked of it: touch theme or plugin files, create or change users, passwords or DNS, flip noindex on its own, edit checkout, cart, account or locked pages, delete content or change prices, email customers or post to social accounts, act on a site other than its own, or spend money without a priced request you started.
That is a different job from fixing it, and worth keeping separate. The free AI crawler check and What AI sees fetch your pages the way GPTBot, ClaudeBot and PerplexityBot fetch them and show exactly what came back, no account needed. A tool that only reports crawl behaviour still leaves the fix to you. The agent's own page check runs the same read, then marks each finding as a switch, words for your approval, or a person's job.
It can fix the access and entity problems that keep a store out of the answer at all, but it cannot write your catalog or promise a specific recommendation. The checks that usually fail on a shop are the same ones above: crawler access on category and product pages, Organization and Product schema that matches the visible price, and an answer-first opening that actually answers "best X for Y." See how ecommerce stores get into AI answers for the shopping-specific version of this. What no tool in this category can do, Reflexa included, is guarantee that an engine names you over a competitor; it can only remove the reasons it would skip you by default.
Four questions, in this order. What surface does it write to, a live page or only a dashboard field? Do you see the exact words or diff before it goes live, or only a summary like "optimized 12 pages"? Is every change reversible on its own? And does it check back afterward that the change actually landed, or does it take its own "applied" status on trust. A vendor that cannot answer the first question plainly is describing a report, not a fix.
None of this requires buying anything, and each step takes a few minutes:
A small category does this: an agent installed on the site itself, not a dashboard beside it. Reflexa Agent is one example that proposes the exact change, shows it in full, and applies it on your click, on WordPress, a coded site's edge file, or a connected GitHub repository.
A tracker counts how often AI engines name your brand and stops there. An SEO suite's AI module adds a suggestion to a to-do list someone still has to action. A site agent writes the exact change, in full, and publishes it once you click apply, with the same hand it used to propose it.
No. It is not a browser agent. It writes through the surface it is installed on: the WordPress plugin API, a coded site's edge file, or a commit on a connected GitHub or GitLab repository. Nothing is simulated by clicking a page.
Ask what surface it writes to (a live page, not a dashboard field), whether you see the exact change before it goes live, whether it is reversible, and whether it checks afterward that the change actually landed rather than reporting "applied" on trust.
It can fix the access and entity problems that keep a store out of the answer at all: crawler access, Organization and Product schema, and an answer-first opening on category pages. It cannot write your product catalog or guarantee a specific recommendation.
The free check reads your site the way ChatGPT, Claude and Perplexity read it, and returns a plan marked switch, words for approval, or a person's job. 3 minutes, evidence included.
The product: Reflexa Agent · how it works · på svenska