Skip to content

Meet Quotable Agent: from AI visibility answer to finished work

Quotable Agent is a chat assistant inside Quotable that answers from your AI visibility data, then does the follow-up work: new articles and rewrites of existing ones, LinkedIn posts, briefs, documents, tables, agents, prompt changes, and more.

There’s no shortage of things an AI visibility dashboard can tell you. ChatGPT stopped naming you for “best CRM for agencies,” a competitor took a citation you used to own, Gemini keeps getting your pricing wrong. The work starts after that: pulling the answers, checking which sources won, deciding whether the drop is real or just nine noisy responses, and turning it into something a teammate can act on.

Today we’re launching Quotable Agent, the AI visibility copilot inside Quotable. It answers from your tracking data, then takes the finding through to the brief, the table, the agent, or the Slack message, and it asks for your approval before anything leaves Quotable.

An answer is only useful if you can check it

Most of the friction in AI visibility work comes from not trusting the number in front of you. A drop shows up, and before anyone writes a brief, someone has to work out how many answers it rests on, whether the engine changed or the prompt panel did, and whether the cited sources mattered at all. So the screenshot gets forwarded, the question bounces between people, and the Monday meeting starts without an answer.

Meet Quotable Agent

Quotable Agent has direct access to almost everything in your workspace, and you can open it wherever the work happens:

  • From Quotable Agent in the sidebar, or with New chat.
  • From the floating widget in the top bar, on any page.
  • From the panel beside a table, when you’re working on rows.
  • From an Action, with Work with agent, so the chat starts on that piece of work.

An empty chat suggests three starting questions from a set of eight you can shuffle through, such as “Why am I not mentioned in ChatGPT?”, “Which sources should I get featured on?”, and “Which model portrays me worst, and why?”

What it knows

Every answer starts from your own data, and the agent calls its tools before it writes a word:

SourceWhat the agent reads
TrackingYour visibility overview, the latest result for each prompt on each engine, your competitors, and the most-cited sources, with each domain labeled as yours, a competitor’s, or independent. Metrics compare periods and filter by engine, topic, market, and prompt type.
ShoppingYour AI Shopping overview and product-level detail.
ChatGPT AdsThe sponsored placements Quotable tracks in ChatGPT answers for your prompts, next to your organic mentions.
Brand contextApproved brand records plus team and personal context from Quotable IQ, ranked by relevance to the question. Locked records stay locked.
Knowledge baseYour company context inside Quotable.
AnalyticsGA4 and Search Console, when they’re connected.
Prompt demandDemand you’ve already measured, and new terms it can measure within your research quota.
AI crawlersReal request logs from AI agents, saved diagnostics, and on-demand audits of robots.txt and individual pages.
WebWeb and news search, up to three searches and three page reads per message.

The web access is fenced in on purpose. The agent only opens URLs you typed, URLs a search returned in that same message, or the homepages of your brand and your tracked competitors. A malicious page can’t instruct it to wander off and send your workspace data somewhere else, because it never gets to choose an arbitrary link to follow.

From answer to finished work

Ask a question and the answer comes with a chart (visibility trend, share of voice, mentions by engine) or a data card, and the written reply doesn’t repeat the numbers already on the card. Ask for the next step, and the agent puts the work where it belongs:

  • Documents, analyses, and plans go to your Library and open in an editable side panel, and you can revise them in the same chat without breaking the charts inside.
  • Full articles and briefs go to your Library as drafts.
  • Agents get built from your description, edited with review control, tested, published, and run (up to two per message, with @), and they’re scheduled only when you ask.
  • Tables get new rows, columns, and agent columns, up to 200 per change, and every write is read back, so you see “All changes confirmed” or exactly what was refused.
  • Tracking gets new prompts within your quota (duplicates are skipped), topic and status changes, and renamed topics, while your response history stays intact.
  • Skills get created or edited in Quotable IQ.

Outside Quotable, every action needs your approval first:

AppWhat the agent can do
SlackSend a message
Gmail and OutlookCreate drafts (it never sends email)
NotionCreate pages
LinearCreate issues
Webflow and ContentfulCreate drafts

If you ask it to use an app you haven’t connected yet, it hands you a connection card.

Write, rewrite, and repurpose content

The agent writes with your tracking data and brand context in front of it, which is the difference between a generic draft and one built around the prompts you’re actually losing. Some of what it can write for you:

  • New articles, drafted from a brief and saved to your Library, built around the prompts and cited sources behind the gap you want to close.
  • Rewrites of articles and pages you already have. Paste the URL or attach the article, and it compares the page with the sources that won the answer, then rewrites the sections that don’t answer the question, adding the missing facts, FAQs, or comparison criteria.
  • LinkedIn posts, newsletters, and threads that turn a finding into something people want to read, with the real numbers and the sample size behind them.
  • Comparison pages, FAQs, and “best X for Y” pages for the shortlists you want to be on.
  • Outreach emails and pitches for the independent publications it recommends, saved as Gmail or Outlook drafts.
  • Battlecards and one-pagers for sales, built on how AI answers compare you with the competitors that beat you.
  • Client and leadership reports, written as a document with the charts inside.

Add a skill to shape the result. Answer-first structure puts the answer at the top of each section, Humanizer strips out the patterns that make text read as machine-written, and your own skills carry your house style from one piece to the next.

What this looks like across teams

The same visibility data leads to very different next steps depending on who picks it up.

SEO: find out why a crawler disappeared

GPTBot requests to your pricing page dropped to zero last week. Ask the agent why, and it checks the AI crawler logs and your saved diagnostics, runs a robots.txt and page audit on that URL, and shows you what it found. Then it drafts a Linear issue for engineering with the evidence attached, which you approve before it’s created.

Content: close a gap you keep losing

A competitor owns the answer for a prompt you track, and the page getting cited is their comparison post. Ask the agent to suggest content to close the gap, answer the two or three questions it asks about audience and format, and the brief lands in your Library. Ask for the full draft and it lands there too, and if you add the Humanizer skill to the message, the draft follows it.

PR and comms: get onto the sources AI trusts

Ask which sources you should be featured on, and the agent ranks the most-cited domains across your prompts, separates yours, your competitors’, and the independent ones, and recommends only the independent reviews, listicles, communities, and media. It can turn the list into an outreach plan in your Library, or a Notion page for whoever runs the outreach.

Leadership: answer “what changed?” before the meeting

Ask what changed in your AI visibility since last week, and you get the comparison by engine with the sample size behind each rate, so a 12-point swing across nine answers reads as directional rather than as news. Ask it to write the summary as a document and post it to Slack, and if you want the same report every Monday, ask it to build an agent and schedule it.

Growth: see who’s paying for the answer

A competitor starts appearing as a sponsored placement on prompts you track in ChatGPT. Ask the agent which prompts carry ads, whose they are, and how they line up with your organic mentions, and it turns the answer into a short recommendation for the paid team, posted to Slack once you approve it.

A simple way to decide what to hand it: anything you’d ask an analyst, ask the agent, and anything you’d normally forward with “can you turn this into…”, let it do.

Skills and Actions

Skills are reusable instructions you call with a slash and the skill’s name. Three come built in: Answer-first structure, Humanizer, and Prompt writer. You can write your own, and import or export any of them as a SKILL.md file, so a team’s house style or review checklist travels between workspaces. The agent never picks a skill on its own, which means the instructions shaping an answer are always ones you chose.

Actions hold the bigger pieces of work. Quotable reads your tracking data and lists them on its own: sources to get mentioned on, content to create or update, and fixes to how AI reads your site, each with its evidence, steps, an owner and a status. Open one with Work with agent and Quotable Agent picks up from that Action, so you don’t re-explain the brief each time.

Get started

Quotable Agent is live in the app now. Open it from the sidebar or with New chat, and every response it writes uses credits from your plan’s allowance.

If you’re not sure where to begin, ask “Why am I not mentioned in ChatGPT?” about the prompt you care most about. You’ll get the answer, the number of responses behind it, and the sources that took the spot, and turning that into a brief is one more message.

Not on Quotable yet? Start a trial and ask your first question against your own data.