Skip to main content
Gaps answers: so what do I go and fix first? It turns a question you are losing into a work item that carries its evidence, and ranks the items by how movable each loss is rather than how large. Three screens work together: Gaps, the queue; Source gap, why you lose one question; and Gap detail, one work item from plan to publication.
Things worth doing to get named more often, each from questions you are losing. Ranked by how much each should move your visibility.

Gaps

What is on it, top to bottom

1

Work items by stage

A strip that counts your items at each stage and filters the list when you click one: Losing prompts, no work item, All work items, Open, In progress, Published, Measuring, Measured and Dismissed.
2

Start here

The highest-ranked item not yet published: what to work on, the question you are losing, and its expected recovery, in points of visibility. Open work item opens it.
3

All work items

The queue, in two groups. Ranked by expected recovery lists items with an estimate, with their Impact, Effort, Owner and Status. Not estimated lists the rest: a claim correction recovers no visibility, and a gap whose sources have too little behind them has no estimate yet. Neither is ranked as a zero. An item whose question you are no longer losing is tagged No longer losing; it is never closed on its own.
4

Losing prompts with no work item

Questions lost by more than their ± band, largest loss first: who they are lost to, by how much, and which sites their answers read. Each opens its Source gap.
If the engines changed how they answer during the window, a line at the top says so. The day is labelled, not removed: a loss is read against the leader on the same answers.

How to read the numbers

  • Impact: “Our estimate of how many percentage points of visibility this could add. It is an informed guess from past answers, not a promise.” It is recalculated each night.
  • Effort: how much work we expect it to take, small, medium or large.
  • Status: where the item is, from open through published to measured.
Each item is one of four kinds: Source gap (get onto a site the answers read), Content gap (a page of your own that answers the question), Claim correction (an engine states something wrong about you) and Technical fix (something on your site stops crawlers reading it).

Source gap

Why you lose one question, on one screen. Open it from a losing prompt on Gaps, or from Why you lose this on Prompt detail.
  • The loss: your visibility against the leader’s, how many points behind you are, and the ± band. A loss larger than its band reads “this is a gap”. A loss within it reads “not a gap”, and you can still raise a work item by hand.
  • Named ahead of you: the brands named more often than yours in the answers to this question, most often first.
  • Choose the sites to work on: the sites these answers read that you could be mentioned on, each with what to do there, how many answers read it, and named more often when read. Some are ticked to start; tick the ones you will work on. Whether you are already listed on each is not checked yet.
    How much more often you are named in answers that read this site than in answers that do not, over the last 30 days. It is a pattern in past answers, not proof that being listed there causes it: only a published change, measured afterwards, shows that.
  • Your own sites and Competitor-owned sites · nothing to do there are shown below, so you can see everything the answers read. A competitor’s site cannot be chosen.
  • Your work item: how many sites you chose, the estimate for them, and Create work item. If every source behind the loss is competitor-owned, the button becomes Brief a page of your own: the hardest kind of gap, where the lever is a page you control.

Gap detail

One work item, from what to do to whether it worked.

The header

The item’s kind, status, effort, owner and impact, the question it comes from (See why you lose it opens the Source gap), and a track showing which stage it has reached. Hand this work item on has two buttons: Copy for your team copies the item as a plain-text message, and, while the item is open, Copy as AI prompt copies it as a prompt you can give an AI assistant to draft the work. If the browser will not copy, the text appears in a box to copy by hand.

What to do

The plan of work: the sites chosen for this item, each with what to do there and the pages on that site the answers named, or A page of your own for a content gap. Evidence, beside it, keeps every site the answers read, grouped as what you can act on, competitor-owned and yours. The ones not chosen stay, so you can see why the item was suggested.

Brief and draft

  • Write a brief says what the answers read and what to write. Write a draft writes a first version of the page. Both cite the evidence for every section, which is shown beside it. Drafts are written by Anthropic’s Claude from the evidence on this page; Anthropic does not train its models on what we send it. Writing one takes up to a minute.
  • Edit a version and Save as version to keep the old one as it was.
  • Approve version approves exactly the text you read. Check each line first: the evidence shows where a sentence came from, not that it is true.
  • Technical fixes are files for your own site, llms.txt and JSON-LD structured data, to view and download. They are written from facts Citino already holds, with no model. They make the page readable; they don’t get you named.
Briefs and drafts count against your plan. See Plans and limits.

Send to your site

Once a draft is approved, you can send it to your own WordPress or Webflow site, after connecting it in Integrations.
1

Send as a draft

Send to WordPress as a draft (or Webflow) first shows what will change on the site: the site, the title, the slug and the text. Nothing is sent until you confirm, and it arrives in your CMS as a draft, not a live page. Later versions go as Update the draft.
2

Publish live, or take it back

Publish live and Unpublish are separate steps, each with its own confirm. Every send is listed under Sent from here, with who sent it.
3

Mark as published

Once the live page answers, Mark as published opens the publish confirm below with today’s date and the page’s address filled in.

Moving the item along

  • Start work moves an open item to in progress; Move back to open undoes it.
  • Effort: set S, M or L.
  • Dismiss asks why, and keeps the reason on the item, so the next reader can tell a wrong gap from one that is just not now. A dismissal cannot be undone here.
  • Mark published asks for the date published and the public URL of the live page, not a draft preview. This only records that you published the change. Citino does not write to your site here; it starts watching for the first answer that reads this page, and an experiment opens (see Experiments). A claim correction is measured on Claims instead.
The owner is shown on the item; choosing one from this screen is not available yet. A step-by-step playbook for each kind of work is not available yet either.

Who can do what

Owners, admins and editors can create, move, dismiss and mark items published, write and approve drafts, and send them to the CMS. Connecting the CMS itself is for an owner or admin. Viewers read everything here, and can copy an item.

Empty and error states

  • No gaps yet: “A gap opens when you lose a prompt in your segments by more than its band.” This shows only when there are no work items and no losing prompts.
  • No work items yet but losing prompts: the losing prompts lead, largest loss first, with a short guide to how a gap closes: choose what to fix, make the change, see whether it worked.
  • This gap is not here: it may belong to another project, or it may have been dismissed.

Experiments

Whether a published change worked.

Sources

Every site the engines read.

Integrations

Connect WordPress or Webflow.

Within noise

Why a loss inside its band is not a gap.