Actions

What to do where AI doesn't mention you, and in what order

Every week Refinea finds the topics where you lose answers and looks for the cause in the data: a missing page, a weak page, third-party sites the engines cite instead of you. For each topic it prepares a plan with agents and, once the work is done, measures the same prompts.

Want to see how AI engines see you today first? Get the free analysis.

  • ChatGPT
  • Gemini
  • Claude
  • Google AI Mode
  • AI Overview
  • Perplexity
  • Copilot
  • DeepSeek
  • Grok
  • Mistral
Nike's Actions: site, pages to rework, new pages and off-site work, with new, planned and in-progress Actions
Refined Marketers
  • Caffeina
  • AIRC
  • Finbooks
  • WeHunt
  • GECOS
  • Grenke
  • Garmont Tactical
  • StoryWalking
  • Omni
  • Seoleader
  • Fabio Antichi
  • Axyon AI
  • Unflat
Named in thesurvey of the tools Italian SEO experts use (2026)
Pre-accelerated byB4i·Participant in TEF Ignition byTech Europe Foundation·Finalist atStartCup Lombardia

The cause, in the data

Every Action says why the topic loses answers and shows the pages and sources where it read that.

How many answers are at stake

The answers at risk on the topic and the search demand behind it.

A plan with agents

The steps, the agent for each step and the estimated credits, before you start.

Measured afterward

The same prompts against the baseline, with the measurement noise threshold.

What you get

Every Action comes with the why, the plan and the measurement.

They come from the week's answers and arrive in order: the site first, then the pages, then the rest.

Four fronts of work

Site, pages to rework, new pages and off-site work: you see right away where the work is.

The four fronts of Nike's Actions: Site 3 fixes, Pages to rework 4 pages, New pages 3 articles, Off-site 3 tasks
The why, with evidence

See which sources the engines cite instead of you and how much weight each one carries.

Ref
Why do assistants cite others on trail running and not nike.com?
I read your data
46% of citations go to publications, blogs and review sites
Top sources: runrepeat.com, correrenellanatura.com, outdoorgearlab.com
24% to Reddit threads and YouTube reviews
On this topic the assistants mostly cite third-party pages, neither yours nor your competitors'. The plan proposes getting into those sources: for each one, what to offer and who to write to.
Engine by engine

On each engine, how often you appear and how often your competitors do, on the same prompts.

The models matrix: the AI Visibility Index of Nike, Adidas, New Balance, ASICS, HOKA and On on ChatGPT, Gemini, Perplexity and AI Overview
The site first

Fixes that apply to the whole site arrive already written, ready to copy.

The site fixes: Article structured data for articles (medium impact), FAQPage markup and llms.txt file (low impact)
Off your site too

When answers draw on publications, portals and Reddit, the Action takes you there with the pitches to send.

Output
Citation Outreach · Marathon shoes
Getting into a ranking
To
Head of testing · runrepeat.com
Page
The best marathon shoes, lab tested
Subject
Alphafly 3 for your marathon shoe ranking

Hello, your ranking is one of the pages AI assistants cite most for marathon shoes, but the Alphafly 3 isn't on it.

If you're planning an update, we can send you a pair for lab testing along with the full spec sheet.

Best regards, the Nike Running team

Correcting a fact
To
Gear editor-in-chief
Page
Marathon shoes: the buying guide
Subject
A spec to update on the Vaporfly 3

Hello, the weight your buying guide lists for the Vaporfly 3 is for the previous version.

We've attached the spec sheet for the current version, in case you'd like to correct it.

Thank you, the Nike Running team

Drafts: you review them and send them yourself.
Ref knows the Action

Inside every Action, ask Ref to explain a step, review it or add one.

‹ BackTrail runningAsk Ref
Ref
Analyze the Action “Get cited where the engines read ‘Trail running’” and prepare a plan to work on it.
I read your data
Reading the Action
On this topic the assistants mostly cite publications, blogs and review sites (46% of citations). I've created three agents to work on it and already started them.
Agents created for this Action
Get cited off your siteCitation OutreachOpen
Write the new pageCreate ArticleOpen
Work on the pageStructured Data GeneratorOpen
The plan

Accept the Action and the agents already know what to do.

Every Action has its steps, and every step has its agent. On trail running, for example, there are three.

  • Free to acceptNo credits are used until you launch a step, and every step shows the estimated credits.
  • Every step has its agentThe pitches for the cited pages, the guide on the unanswered questions, the structured data for the existing page.
  • Your team approvesAgents stop at the outline and the drafts until someone on the team says yes.

Explore the agents

‹ BackTrail runningAsk Ref
Agents created for this Action
Get cited off your siteCitation OutreachOpen
Write the new pageCreate ArticleOpen
Work on the pageStructured Data GeneratorOpen
The measurement

The measurement is what closes the Action.

Every day Refinea runs the topic's prompts again and compares the AI Visibility Index with the baseline, counting only changes larger than the measurement noise.

  • A stated goalRaise the AI Visibility Index on the topic's prompts above the baseline, within a set period.
  • The same promptsThe measurement uses the same prompts and the same window as the baseline.
  • Improved or unresolvedAbove the noise threshold the Action closes as improved. Within the noise it stays unresolved, even if you mark it as done.
The AI Visibility Index on trail running: 40.2%, baseline 39.6%, goal above the baseline within 30 days
How it works

New Actions, every week.

  1. 1

    The Actions arrive

    The week's analysis opens an Action for every problem it finds on the site and across topics.

  2. 2

    Choose what to work on

    Accept the ones that matter and dismiss the others with a reason: already done, not relevant, not now.

  3. 3

    Work and measure

    Launch the steps, approve what the agents prepare, then Refinea measures the same prompts.

“GEO stopped being guesswork: Refinea turned it into a to-do list.”
Marcello RicciMarketing at FinbooksFinbooks

Put agents to work on everything AI reads about you.

With Pro, we set up Refinea with you on day one. Agents prepare the work, and your team approves it before anything goes out.

Want to see how AI engines see you today first? Get the free analysis.

Frequently asked questions about Actions

What are Actions?

They're the topics where your brand loses answers in AI engines, each with the cause read from the data, the answers at stake, the plan with agents and the measurement after the work.

How often do they arrive?

Every week. The analysis of the week's answers opens an Action for every problem it finds.

Does accepting an Action cost credits?

No. Credits are used only when you launch a step, and every step shows how many it needs before you start.

Are Actions only about the site?

No. Beyond the site and its pages, they take you where answers read about you: publications, portals, review sites, Reddit and YouTube.

How do I know if the work paid off?

Refinea measures the same prompts before and after and tells you whether the AI Visibility Index on the topic rose beyond the measurement noise. It shows you what changed, without claiming credit the data doesn't prove.

Can I dismiss an Action?

Yes, with a reason (already done, not relevant, not now, topic covered), and you can restore it whenever you want.

Can I take them out of Refinea?

Yes. You can export them to a file with type, topic, impact, reason and the pages involved.