Skills/Agents/SEO + LLM Dashboard
Agent

SEO + LLM Dashboard

/seo-llm-dashboard · by Victor Shulga

One page: how you are found in Google and what ChatGPT, Gemini, Claude and Perplexity answer about you. One skill does it all: the measurement, the scoring and the page. Every question and each model's answer: whether it named you, whom it recommended instead, which sites it cited.

npx skills add victor-shulga/seo-llm-dashboard/skills/seo-llm-dashboard

How to set it up: 5 steps

An AI-only start needs steps 1, 4 and 5. Steps 2 and 3 add Google data and can wait. Claude walks you through every step, and you paste keys with a terminal command, never into the chat.

1Step 1. Connect DataForSEOrequired

Sign up at app.dataforseo.com. A new account gets $1 for testing: enough to check the key, not for a measurement. For a measurement top up by card, the minimum payment is $50 and the balance does not expire. Current terms are on the DataForSEO pricing page.

The API login and password are in your DataForSEO dashboard under API Access. It is not your dashboard password. Copy the string from the «Base64 Format» field and save it on your computer with one command (Mac):

mkdir -p ~/.config/dataforseo && pbpaste > ~/.config/dataforseo/key && chmod 600 ~/.config/dataforseo/key

The key sits in a file outside the project and outside git, never in the chat. Then the free check: Claude runs avr.py check. «DataForSEO OK» with your balance means it works. «401» means the wrong string was copied, usually the dashboard password.

A measurement costs about 4 cents per model answer: 27 questions × 4 models = 108 answers ≈ $4. Before collecting, the skill shows the sum and waits for your yes.

2Step 2. Search Consoleoptional

The dashboard reads Google data through OpenSEO, a free open-source app installed on your own Cloudflare account; the free plan is enough. Claude deploys it with you and locks it to your email only.

Then Google Cloud: turn on the Search Console API, create a web client, publish the app (unpublished, access dies after 7 days) and connect the site with the Connect button in OpenSEO. Plan an hour or two. Done when Claude shows clicks and impressions for 28 days.

3Step 3. Positions against competitorsoptional

Needs step 2. In OpenSEO add your domain with 8–15 queries you should be found for, and each of 3–5 competitors with the same list.

Positions are checked weekly, about $0.08 per domain from the same DataForSEO balance.

4Step 4. First measurement and the dashboardrequired

Tell Claude: «set up the SEO LLM dashboard for your-domain.com». In one message it asks for the brand, market and buyer languages, 4 topics and 3–5 competitors, writes the questions, shows the sum and after your yes collects the answers of four models, 5–15 minutes.

Then it reads every answer that names you, computes the score, works out why each lost question was lost and writes the «what to do» recommendations: each with a number from the measurement, effort and priority. With Google it adds 1–3 search ones when numbers back them. All of it lands in one HTML file: open it from disk or publish it.

5Step 5. Monthly refreshrecommended

Once a month the same questions go to the same models, and the chart shows how your score and your competitors' scores move. Recommendations are rewritten after every measurement. Claude sets up a scheduled task with a budget cap you name.

With Google connected a weekly refresh of clicks and positions is added, free. Treat a change under 1 point between months as noise.

The full guide ships inside the skill, file SETUP.uk.md (Ukrainian).

What it does

Asks buyer questions to four models with live web search through DataForSEO and turns the answers into an AI visibility score: total, per question language, per model and next to competitors, plus the sites the models cite and how the score moves from one measurement to the next. Below sits the prompt table: a row is a buyer question, the columns are ChatGPT, Gemini, Claude and Perplexity, each cell says 1st, yes or a dash. Click a row to read all four answers. Connect Search Console and OpenSEO and clicks, impressions, queries and keyword positions against competitors appear on top. Without them the page is built from the AI part alone and the other sections hide themselves. The result is one HTML file: open it on your computer or publish it. The design is fixed: Claude only fills in your data and writes the recommendations, and a script checks after each build that every block is in place.

Input and output

Input
A DataForSEO key, brand and domain, 3 to 5 competitors, 4 topics phrased the way a buyer asks, market and languages. For the full version also Search Console and OpenSEO.
Output
One HTML page: AI visibility score by language and model, competitors, cited sites, every prompt with the answers of four models, the reason behind each lost question and recommendations after every measurement. The full version adds clicks, queries and Google positions.

What it costs

Key features

How to install

1From the terminalfastest

Node 18+ required. The command installs the skill into the current project folder:

npx skills add victor-shulga/seo-llm-dashboard/skills/seo-llm-dashboard

To keep the skill at hand in every project, add -g:

npx skills add -g victor-shulga/seo-llm-dashboard/skills/seo-llm-dashboard
2Download the ZIPno terminal needed

In Claude Code: unzip the folder seo-llm-dashboard into ~/.claude/skills/ and restart the app. In Claude Desktop or on claude.ai: Settings → Capabilities → Skills → upload the same archive.

Download the ZIP · 121 KB
3Manually from GitHub

Open the skill's folder on GitHub and save the files. The main file is SKILL.md, and the rest is pulled in from it.

Check it works: type /seo-llm-dashboard in the chat, and the skill should answer. If it does not: skills are read at session start, so restart Claude.

Related skillsAll skills →