AI Agents
A team of agents working your AI visibility.
Twelve specialized agents that learn your business, find gaps, surface opportunities, watch for changes, and report the visits and conversions AI engines send you, from your GA4 data.
The Business Profiler reads your website the moment you add a project: crawling your homepage and the key pages it links to, to learn what you sell, who you serve, and where. From that it derives your business type, target markets, and focus keywords, and proposes a starting set of prompts to track. Everything is staged for your review: nothing about your project changes until you approve it.
- Crawls your homepage and the key pages it links to, to learn what you do
- Derives your business type, target countries, and focus keywords
- Proposes a day-one set of prompts to track, all staged for your review
Example
For example, it might profile a Denver cake shop as a local Food & Beverage brand targeting “custom birthday cakes near me,” and stage a starter set of prompts to track.
Prompt Discovery keeps your tracking list sharp. It generates buyer-intent prompts from your domain, business profile, competitors, and the pages it crawls (spread across awareness, consideration, and decision), then ranks them by opportunity, dedupes against what you already track and recently saw, and surfaces a short “must-track” set you accept with one click. On Starter, Growth and Pro, the project owner can turn on its weekly pass in project settings, and it then proposes new prompts every week; on Growth and Pro, owners and editors can also ask it for more.
- Generates buyer-intent prompts across awareness, consideration, and decision
- Ranks by opportunity and dedupes against the prompts you already track
- Refreshes weekly once the project owner turns it on, with more on demand on Growth and Pro; accept the best with one click
Example
For example, it might propose “best CRM for startups” and “HubSpot alternatives for small teams” as top must-track prompts, ranked ahead of a dozen lower-value variants.
Being mentioned isn’t the same as being recommended. The Sentiment & Framing agent reads every AI answer that cites you and classifies how you’re positioned (the top recommendation, one option among several, a passing mention, or something the AI actively warns against), backing every verdict with the exact line from the AI’s answer. On Starter, Growth and Pro it also says why: it catches when an engine recommends a competitor over you and names the reason (pricing, features, integrations, support, security). It turns “we got mentioned” into “recommended as the budget option, but passed over for Competitor X on pricing” and feeds that signal to every other agent.
- Classifies each citation by role (top recommendation, alternative, passing mention, or warned-against), not just a mention count
- Names the reason behind the framing on Starter, Growth and Pro: pricing, features, integrations, support, security, and more
- Flags when an engine recommends a competitor over you, with the reason it gave on Starter, Growth and Pro
- Backs every verdict with the exact excerpt from the AI’s answer, and feeds tone into your gaps, alerts, and content
Example
For example, it might reveal that ChatGPT recommends Competitor X over you for “best CRM for startups,” specifically on pricing, and quote the line from the answer that proves it, so you know exactly what to fix.
The Visibility Watchdog runs every day and tells you the moment something material moves, not at the end of the quarter. It goes past “you dropped”: it names the prompts that drove the change, attributes it to the engine behind it (a ChatGPT shift reads differently from a Perplexity one), and tells you whether it’s you specifically or a category-wide swing every brand felt. Still cited as often but slipped from #1 to #8? It catches that too. Findings appear in the alerts feed and in-app notifications. On Starter, Growth and Pro, owners can choose daily or weekly digests, select alert types, or turn alert email off; each email includes up to five recent highlights. Saved domain or exact-URL watchlists can also subscribe to citation gains and losses through team notifications and, on those plans, owner email, with saved source counts, dates and recorded-channel evidence.
- Catches both overall and per-prompt drops, and names the prompts that moved
- Attributes each change to the engine behind it (ChatGPT vs Perplexity vs the rest)
- Tells you if it’s you-specific or a category-wide shift, so you don’t chase engine noise
- Rank-slide detection and a full alerts feed, with daily or weekly owner alert email on Starter, Growth and Pro
Example
For example, it might tell you your citation rate for “best CRM for startups” fell 42% this week (driven almost entirely by ChatGPT, while competitors held steady), and link you straight to the affected prompt.
When AI cites a competitor instead of you, the Citation Gap Finder reads the winning page and names the exact change that would close the gap: a comparison table, an FAQ, missing statistics, weaker schema, or stale content. It doesn’t hand you an undifferentiated to-do list: it builds a queue ranked by impact, effort, and confidence, so you always know the single highest-value fix to make next. Because one page often answers many prompts, a single change can win back a whole cluster at once, and after you ship it, the agent tracks your citation rate against the recorded baseline, which shows what followed rather than what the change caused.
- Builds a ranked, evidence-backed queue: every gap ties to a real page an AI engine cites, scored by impact
- Names the exact fix (comparison table, FAQ, statistics, schema, or a content refresh), not “write more”
- Clusters related prompts, so one page fix can lift many at once
- Measures what followed: tracks your citation rate at 7, 14, and 30 days after each change, against the recorded baseline
Example
For example, it might rank a single fix, add a comparison table and a pricing FAQ to one page, as your top opportunity, because that one change could win back 9 prompts where competitors are cited across ChatGPT and Perplexity.
The Article Topic Planner scores every tracked prompt by content opportunity, weighing competitor citations against your own brand share, and generates a ranked, weekly slate of article topics most likely to earn AI citations.
- Scores prompt opportunities by competitor citations vs. your brand share
- Generates a fresh, ranked set of article topics every week
- Focuses your content team on the topics with real upside
Example
For example, a top pick might be “How to migrate from Salesforce”: high competitor citations, little brand share of your own.
The Article Writer starts where most AI writers won’t: it decides whether a page should exist for a prompt at all. Before it drafts a word, it builds a brief. It reads the pages AI engines actually cite for that prompt alongside your own site, then recommends the real move: a new article, an update to a page you already have, a comparison or FAQ block, or holding off. When a new article is the right asset, it drafts from the cited sources, labels every claim as verified, unsupported, or a proof point for you to supply, scores how citation-ready the draft is, and hands you a publish checklist covering the slug, metadata, JSON-LD schema, and internal links. Once you publish the page, it ties back to your prompt monitoring to see whether it earned the citation.
- Builds a brief first, and may tell you to update an existing page, add an FAQ, or hold off, instead of writing a post you don’t need
- Drafts from the sources AI actually cites for the prompt, labeling every claim as verified, unsupported, or a proof point for you to supply
- Scores citation-readiness and hands you a publish checklist: slug, metadata, JSON-LD schema, and internal links
- Ties the published page back to your prompt monitoring, so you can see whether it earned the citation
Example
For example, for the prompt “How to migrate from Salesforce” it might recommend updating your existing docs page rather than writing anything new. When a new article is the right call, it drafts one grounded in the pages ChatGPT cites today, flags the two statistics you need to supply, and gives you the slug, schema, and internal links to publish it.
The On-Page Optimization agent turns a known citation gap into a copy-ready page edit. It reads your live page alongside the competitor pages AI engines actually cite for your prompts, then hands back a ranked worklist: which page to fix first, the exact change to make, and the evidence behind it. Every recommendation ships paste-ready (a Markdown checklist, a CMS-safe HTML snippet, or a JSON-LD schema block), and every retained change carries at least one quote matched to the saved source text, though that check does not verify every claim in the draft. Regulated medical, legal, and financial claims are flagged for human review, nothing is ever applied automatically, and once you mark a change done it re-runs the affected prompts and compares your citation rate with the recorded baseline, without establishing that the edit caused the difference.
- Ranks a worklist by impact, effort, and confidence, so you always know the single highest-value page to fix next
- Names the exact change (FAQ schema, comparison table, a missing entity or statistic, a freshness update, or restructured headings), with an estimated time to apply
- Ships copy-ready output (Markdown, CMS-safe HTML, and paste-in JSON-LD), each retained change quote-matched to the page AI actually cites, with the draft left for you to review
- Measures what followed: after you mark a change applied, it re-runs the prompts and compares your citation rate with the recorded baseline
Example
For example, it might rank one product page as your top fix, hand you a paste-ready FAQ schema block and a drafted comparison table, the exact elements the competitor page ChatGPT cites already has, and flag it as a 15-minute change. After you apply it, it watches whether the prompts you were losing start citing you.
llms.txt is the emerging standard for telling AI crawlers what your site is and which pages matter most. The llms.txt Advisor fetches your live file and grades it against the llmstxt.org spec, a clear pass, warn, or fail on every rule, then explains what to fix in plain language and generates a compliant llms.txt built around your site’s key pages, ready to publish. It re-audits every week once the project owner turns that on, and on demand up to once a day, so your file stays correct as your site changes.
- Fetches your live llms.txt and grades it against the llmstxt.org spec, pass, warn, or fail on every rule
- Explains what to fix and why in plain language, not raw validator errors
- Generates a compliant llms.txt built around your key pages, ready to copy and publish
- Re-audits weekly once the project owner turns it on, and on demand up to once a day, so the file stays correct as your site evolves
Example
For example, it might flag that your llms.txt is missing the required H1 and summary blockquote, then hand you a corrected file listing your product, pricing, and docs pages in the exact format the spec expects.
The Search Plan turns your Search Console data and your tracked prompts into a plan. Every week, and whenever you rebuild it, an AI model groups the Google searches that find your site and the prompts you track into topics, using only your project’s own data: your searches, prompts, pages, keywords and business profile, with no web search. For each topic it shows how you do on Google (impressions, clicks and your best position) next to how often AI answers cite you, gives the topic a state such as “On Google page one, low AI citations”, and, where there is one, the next step: track a prompt, write a page, or improve the page that ranks. The model groups and names the topics, judges their relevance and writes the reason for each, but produces no number: Google and AI numbers come from your own data, and monthly search volume for the topics about your business comes from DataForSEO’s Google Ads data.
- Groups your Google searches and tracked prompts into topics weekly and on demand, from your project’s own data
- Shows each topic’s Google numbers beside how often AI answers cite you, with a state and, where there is one, the next step
- Adds monthly search volume from DataForSEO (Google Ads data) to the topics about your business
Example
For example, it might show that a “Recurring invoices for retainer clients” topic has a best Google position of 4.7 while AI answers cite you in only 5 of 40, and point you at improving that ranking page for AI.
The Revenue Attribution agent imports your Google Analytics 4 data every day, keeps the sessions that arrived from an AI engine’s own referrer, and reports them by engine and by landing page, with the conversions and revenue GA4 recorded for them. It answers the question every visibility report raises: when AI answers send people to your site, where do they land and what do they do next? Visits are credited to an engine and a page, not to a single citation, because the import reads the referring engine and the landing page, not the answer behind the click.
- Imports daily GA4 engagement and revenue metrics
- Splits AI-referred sessions, conversions and revenue by AI engine
- Shows the landing pages AI engines send visitors to, and what those visits convert into
Example
For example, it might show that AI-referred visitors convert at a noticeably higher rate than your site average.
The Growth Advisor turns your raw metrics into a decision. Every Monday once the project owner turns it on, and whenever you ask (up to once a day per project), it pulls your Google Search Console and GA4 performance, runs live research across your market and competitors, and writes a narrative briefing tied to the business goal you set: what moved, why it moved, and the specific next moves to make. It is the difference between a dashboard full of numbers and a plan you can forward to your team.
- Synthesizes your Search Console and GA4 performance with live web research into one narrative briefing, weekly or on request
- Reads every trend against the business goal you set, so the advice targets your objective, not generic best practices
- Names what moved, why it matters, and the concrete next moves worth making this week
Example
For example, it might tie a two-week slide in AI-referred signups to a competitor’s new comparison page, name the exact prompts you’re losing, and lay out the two pages to publish this month to win them back.
Put these agents to work for your brand.
Let DiscoveredBy’s agents monitor, diagnose, and improve your AI search visibility for you.
Start free