Clicky skills

A skill points Clicky at one job for a single answer. Instead of a short reply to a short question, you get a structured run: Clicky reads a specific set of your data, follows a method built for that job, and answers in a fixed shape.

Running a skill

Type / in the message box and pick one, or open Skills from the conversation list on the Agent page. Either way the box fills with the skill's opening question — send it as it stands, or edit it first to narrow the run to a topic, a competitor, or a set of pages.

The four skills are deliberately separate. Each one stays in its lane, so a visibility audit will not drift into an outreach list and an action plan will not turn into page edits. Run them in order when you are starting from scratch.

Visibility Audit

Where you stand in AI answers, prompt by prompt, and why.

Clicky sorts every prompt you track into three states, because the fix differs for each:

• Cited — one of your own pages was used as a source in the answer.
• Mentioned but not cited — the answer named your brand with no page of yours behind it.
• Absent — the answer did neither.

It then builds a citation-source map: how much of your citation share comes from your own pages versus competitor domains versus independent publishers. That split is what separates an on-site problem from an off-site one — if your own pages supply a minority, the gap is not something new content on your site will close.

Each weak prompt gets a diagnosis with the evidence behind it: nothing to retrieve at all, ranking but not cited, cited only through other people's pages, or visibility that was there and went.

Use it when you want the honest picture before deciding anything, or when a number moved and you want to know which prompts moved it.

Source Action Plan

Where to act next, ranked, from the sources AI already cites for your prompts.

Clicky reads the sources cited for your tracked prompts and splits them two ways: pages worth copying, and pages worth getting onto. Both come from the same list — the second is the first filtered to where your brand does not appear at all.

Candidates are ranked by how many prompts the page appears in, how many topics it spans, and how often it is cited. A page appearing across hundreds of prompts is a different class of target from one appearing in twenty.

The plan also checks three gates before anything else pays off: whether AI crawlers can actually reach your pages, whether your site carries the evidence that you are a real company, and whether there are queries where you already rank well and are still missing from the citations. That last set is cheaper to fix than anything new.

You get a ranked queue. Each row names the source or page, the figures that put it there, and the one action to take.

Use it when you know you are behind and need to decide what to do first.

Citation Shape Audit

How your own pages are written, judged against the shape AI assistants actually pull from.

Retrieval works at the passage level — a single section of a page can be cited while the rest is ignored. This run reads your pages and checks them against what that implies: an opening answer of two or three sentences, self-contained sections that survive being quoted alone, headings phrased as the question a buyer would type, a comparison table, a visible author and last-updated date.

It reads a page that is already winning the same query too, so the verdict compares against what models are lifting today rather than against a generic checklist.

You get a verdict per page: the rules it breaks and the rewrite that fixes each one — the heading to change, the opener to cut, the table to add, and the exact sentence to replace.

Use it when you have pages that should be getting cited and are not.

Community Placement

The Reddit, LinkedIn, YouTube and other threads worth taking part in, and what to say.

Clicky first checks whether community sources are actually cited in your category — a platform can be most of the citations in one category and a rounding error in the next, and only your own account settles which one you are in.

It then ranks the threads AIclicks has already scraped and analysed, using how often AI models cite each one, the prompts and topics it covers, and the thread's own engagement figures. Dead and removed threads are dropped; threads not yet analysed are called out as unread rather than ranked badly.

You get a ranked list, and for each row the comment to leave — written out in full, in a register that will not read as brand copy. The plan comes with the constraints that make this work: comment on threads that are already cited rather than starting new ones, no link in a top-level comment, never the same comment twice, real accounts, and community rules respected rather than worked around.

It also sizes the plan honestly. Roughly sixty percent of comments survive moderation, first movement in AI answers takes four to eight weeks of consistent mentions, and the durable version compounds over six months or more.

Use it when the audit or the action plan says your gap is off-site.

Next steps

• Clicky — what Clicky is, where to find it, and how it reads your filters.
• Use sources — the manual version of the action plan.
• Engage — the Engage tab, where the community threads come from.