Question 1

Can I track changes to brand mentions inside cited source pages?

Accepted Answer

Yes. Owners and editors can save source-mention snapshots on every plan; all current members can read them. The latest 30 per project freeze names, aliases, source evidence and saved AI/team assessments for up to 500 cited third-party pages in a 7, 28 or 90 day window. Compare literal matches or reviewed brand meaning across compatible snapshots. Explicit AI review on standard Starter and above helps distinguish real references from unrelated name matches using quoted evidence, with an uncertain state and owner/editor corrections. Include-all, include-any and exclude filters apply together; missing or uncertain evidence never proves absence. Changed aliases, missing text or populations, changed destinations, and incompatible review methods are not comparable. A changed judgment on identical text is not a gained or lost page mention. Saving uses existing evidence; it does not fetch pages or run AI.

Question 2

Can I choose how often alert and report emails arrive?

Accepted Answer

Yes. Project owners can choose daily, weekly or no alert email, and select any of nine alert categories: the original four (citation rate drops, competitor citation surges, prompt citation drops and citation rank slides) plus visibility and share of voice shifts (your own Visibility rising or falling, and a competitor’s share of voice rising), emerging brand (a brand that starts being named in your answers), sentiment shift (now also covering a competitor’s negative sentiment moving and a competitor’s weakness reason rising), competitor-only social citation and fact contradiction (an AI answer contradicting one of your approved facts). The weakness-reason checks, the social-citation category and fact contradictions only ever have anything to send on a plan with Brand reasons, Social sources or Fact check respectively. Owners can also independently turn generated weekly-report email on or off. Daily sends at most once per UTC day; weekly becomes due Monday UTC and sends on the next watchdog run. Digests include selected saved alerts from the last seven days since the previous successful digest, excluding anything already dismissed, with up to five recent highlights and the total count. Empty digests are skipped. Existing owner plan gates apply. In-app alerts, saved reports and other product emails are unaffected; this does not add Slack or webhook delivery.

Question 3

Can I acknowledge, dismiss, snooze or get an explanation for an alert?

Accepted Answer

Yes. Owners and editors can set any alert to open, acknowledged or dismissed and attach a short shared team note; viewers cannot. Status and note are one shared reaction per alert, visible to the whole project, not a personal setting, and are replaced by whatever is sent each time. The snooze is separate: leave it out to keep the alert’s current snooze, clear it explicitly, or choose 7, 14 or 30 days to (re)snooze from today. Reopening an alert while its snooze is still active is stored as acknowledged instead, since an alert cannot stay unactioned while also being silenced, and a snoozed match keeps suppressing a repeat of that same alert for as long as the snooze lasts. On Starter and above, owners and editors can open an alert and request a plain-language reading of its own saved evidence: title, message, delta, windows and quotes already on the page, with instructions telling the model to describe what changed and never guess why; viewers can read an existing reading but not request a new one. The reading is a summary of up to 600 characters plus up to 3 observations of up to 300 characters each, cached per alert while it stays among the project’s newest 500 narrations, limited to one request per project per minute, and treated as an AI reading of the saved evidence that can be wrong and never establishes a cause. Every alert is detected and shown on every plan regardless of whether its owner can request a reading of it.

Question 4

Can I run my tracked prompts as a specific kind of buyer?

Accepted Answer

Yes, on Starter and above. Settings → Personas lets owners and editors name a buyer persona with a short description; a tracked prompt can then run as General, one or more personas, or several together. Each country, audience and language combination uses one prompt slot, the same as adding a country does today. A persona run adds one audience block naming the persona to the message sent to each chat engine, Perplexity included; it is not a logged-in consumer session. Google AI Overviews, AI Mode, ChatGPT and Gemini receive only the prompt text, so persona variants do not run there. Archiving a persona frees its prompt slots and keeps its history; restoring it does not bring those targets back. Personas have their own on-demand prompt-suggestion request. Sentiment trends and Brand reasons each offer a by-persona breakdown, and Explorer can break any of its metrics down by persona. How many personas can be active at once depends on your plan; see Pricing.

Question 5

Can I run a tracked prompt in another language, or restrict which engines run it?

Accepted Answer

Yes, on every plan. Add or edit a prompt and name one or more of 48 supported languages: the prompt text itself is never translated, the engine is simply asked to write its whole answer in that language, and every chat engine, Perplexity included, gets that instruction the same way. Google AI Overviews, AI Mode, ChatGPT and Gemini receive the language as a setting of the request instead, and support every language but Chinese. Each country, audience and language combination uses one prompt slot, the same as a country or a persona does; a prompt tops out at 100 active variants, and every way of adding one (creating or editing a prompt, bulk-applying a template, CSV/Excel import, accepting a suggested or discovered prompt) is refused once it would go over. Editing a prompt or bulk-applying a template to one is the one exception that can keep, but never grow, a grid already larger than 100 from before this release. You can also restrict a prompt to a chosen subset of engines, which costs no prompt slots and only changes how many answers it produces. A server-side preview shows the new slot cost and a monthly estimate (runs, when a Google engine is included, since Google does not show an AI answer on every run) before you save. Save a combination of countries, audience, languages and engines as a named template, up to 20 per project, to reuse on the Add prompt form or bulk-apply to several tracked prompts at once.

Question 6

Can I find tracked prompts I could pause to free prompt slots?

Accepted Answer

Yes, on every plan, through Prompts → Coverage. It compares the last 28 days of answers to the prompts running now, counting the domains their answers cite and the brands named in analysed answers, and recommends a set you could pause together while everything those prompts see regularly (in at least 10% of their answers, on at least 2 days) is still seen by a prompt you keep. A prompt needs analysed answers on at least 14 days before it is judged. Prompts chosen for fact checks, named by an open or still-measuring optimization, or linked to a citation gap you accepted, are working on or are measuring are never recommended, and neither is the last prompt measuring a country, city, persona, language or engine. Any member can preview a selection, which lists the slots it frees, what it would stop observing (including sources and brands it sees only occasionally, which are not guaranteed) and any value it would stop measuring; owners and editors can then pause it. Paused prompts keep their history, a fix or citation gap whose prompts have no answers stays unsettled instead of being judged on zero answers, and resuming a prompt needs free slots. The comparison pools engines, locations, personas and languages.

Question 7

Can I review automatic prompt classifications?

Accepted Answer

Yes. With project-owner plan access (standard Starter, Growth and Pro), owners and editors can request AI suggestions for up to 25 existing prompts, then edit and review intent, buyer stage, theme and branded or unbranded classification before saving. Branded means naming a tracked own or competitor brand or product. Requests are on demand, limited to one batch per project per minute, and use the configured model. Previews expire after 24 hours and refuse stale prompt or brand context. Manual controls, bulk editing, a branding filter and prompt exports are also available.

Question 8

Can I invite a teammate before they have an account?

Accepted Answer

Yes. Project owners can create a seven-day invitation link for an email address and role, copy it, and share it themselves. The recipient must sign in with that exact address, review the project and role, and accept. Normal sign-in policies apply. Owners can revoke pending links. Invitations do not reserve seats; capacity and the inviter’s current owner access are checked on acceptance. Creating a link does not send an email.

Question 9

Can I manage outreach to sources that AI answers cite?

Accepted Answer

Yes. Earned sources lets owners and editors save cited third-party pages with evidence, add a prospect by hand before it is ever cited, assign active project teammates, keep notes and follow-up dates, and track outreach or a reported placement. All project members can read on all plans. Discovery excludes your domain and active tracked competitors. Your team handles outreach externally; DiscoveredBy does not send messages or arrange placements. After you report a placement on the saved page, compare its observed citation rate in equal 7, 28 or 90 day windows, with separate recorded channels and explicit counts. A rate change does not verify the placement or establish that outreach caused it. Partner integrations are not included.

Question 10

Can I see which pages AI engines found but did not cite?

Accepted Answer

Yes. Retrieved vs cited shows, for one recorded collection channel at a time, how often each domain or URL was returned by the engine while it answered your prompts and how often the answer text attributed it, over 7, 28 or 90 days ending yesterday. A source counts as attributed only when a passage or text position was recorded with the credit, so pages an engine merely listed do not count; that is why these numbers can be lower than the Citations page. Claude and Perplexity report their search results and Grok reports the pages it opened and listed, so all three are compared. Gemini API reports only its grounding sources, which it nearly always credits, so it is not compared; Google AI Overviews, Google AI Mode and Gemini report only the sources their answer cites, and the pages ChatGPT’s searches found are stored but not compared yet, so none of those four is compared either. Returned sources are what an engine reported, not every page it read. Select a URL to see the answers that returned or attributed it, up to three recorded passages per answer, tracked brand matches and our full saved copy of the page, which may differ from what the engine read. All project members can read it on all plans; nothing is fetched or sent to a model when you open it.

Question 11

Can I see which answers show shopping, local businesses, ads or tables?

Accepted Answer

Yes, on every plan. Answer features reports, per engine, how often its answers held shopping, local businesses, ads, video, images or tables, and how often they used web search, as rates over the answers we checked; the prompt page shows a card of them and a chip on each run, the Explorer has seven Answer features metrics, and, with the project owner’s data-export access, an export lists every answer. Shopping, local businesses, ads, video, images and tables are counted from DataForSEO’s copy of the Google AI Overviews, ChatGPT or Gemini response, from its structured parts only: we never read the answer text, and the only links we read are Google’s product-viewer links and YouTube video links inside an AI Overview, and the Google product and Google Maps place links among Gemini’s sources. Google AI Mode draws its images and product cards inside its answer text, so it reports no feature we can count, and AI Overview images are not reported. Web search is what each engine recorded: its recorded searches or returned sources for the API engines and ChatGPT, whether the answer showed sources for Gemini; it is blank for Google AI Overviews and AI Mode, whose AI answers are part of Google Search. In our live checks on 2026-09-27 and 2026-09-28 we saw tables, products and local businesses in ChatGPT answers, product-viewer links in one AI Overview and links to YouTube videos in three of the four AI Overviews we checked, but no ads or images, and no shopping, table or video part from Google; those are counted from the data provider’s documentation. In Gemini answers we saw tables and local businesses, and shopping only in one answer for the United States whose product links were listed for Malaysia; an answer whose product or place links are listed for another country is left out of that rate, and we have never seen images there. An AI Overview counts as having video when it holds a video part or links a YouTube video. An engine that does not report a feature shows no value, never 0%.

Question 12

Can I compare which cited sources are gaining or losing visibility?

Accepted Answer

Yes. Source comparison shows domain citation rates across adjacent 7, 28 or 90 day periods, ending on a date before today that you choose. Dates follow the collection server’s calendar, and elapsed days can still receive late collection results. Each comparison keeps the recorded provider, platform, surface, collection method and model separate. It includes sources cited in either period, with counts, denominators and percentage-point changes. Owned-source share separately divides your domain’s appearances by all domain appearances; each domain counts once per output. Missing periods are unavailable, zero means no saved appearance in collected outputs, and small samples are provisional. Prompt and country mixes can change, so movements do not establish cause. All project members can read it on all plans. Owners and editors can save up to 20 project watchlists of up to 100 cited domains or exact URLs each. URL lists compare each saved page separately. Watchlists filter source rows while keeping output denominators and owned-source share unchanged; saved members with no citations in the periods remain visible. Lists use current membership. Owners and editors can opt each list into team in-app notices and owner email for citation gains and losses across adjacent seven-day periods ending yesterday. Each recorded channel needs 30 completed outputs and five observed dates in each period; repeated source/direction events are suppressed for seven days. Gains mean appearing after zero saved citations, and losses mean disappearing after being cited. Owner email follows existing plan access and daily/weekly/off cadence. Editing, unsubscribing or deleting a list cancels its queued email. Expand Watchlist evidence to see prompt reach, consistency and influence for the selected current period and channel, together with literal brand matches in current saved page text. Fetch dates, unknown scrapes and bounded text coverage are explicit.

Question 13

Can I investigate negative sentiment by topic or reason?

Accepted Answer

Yes. Sentiment trends filters saved results by current prompt intent, buyer stage, theme and persona, within one collection channel, and offers a by-persona breakdown for one brand family. Use Review reputation risk to inspect your brand’s negative mentions or saved Warning or caveat and Not recommended roles together, counting each mention record once. Filter by brand, sentiment and role without changing the comparison totals. These are saved classifications to inspect, not verified harm, a risk score or new alerts. A separate citation framing view groups saved citation reasons such as pricing, features and support, with quotes, rationale and confidence to inspect. Citation reasons describe cited domains and are not explanations assigned to brand mentions. For reasons given about each brand itself, Brand reasons (Starter, Growth and Pro) counts, per brand family, how many answers gave pricing, features, support or ten other reasons as a strength or a weakness, with the quote from the answer behind each count, grouped individually or folded into six fixed categories; answers analysed before that feature are shown as not analysed rather than re-analysed. A separate Compare vs. competitor screen restricts the same kind of comparison, sentiment, roles and reasons, to only the answers naming both your brand and one chosen competitor. Objections (Starter, Growth and Pro) is a separate weekly study, not a view of your tracked answers: it asks each chat engine on your plan, and ChatGPT (not Google AI Overviews, AI Mode or Gemini), directly why a buyer might not choose your brand and each active tracked competitor, groups the objections by meaning with an AI model, and scores each by how early the engines list it, with the quote behind it. The question invites objections, so most brands get some; compare them across brands and weeks rather than reading them as a verdict. These model classifications do not establish what caused a change; editing a prompt classification also changes its historical segment.

Question 14

Can I see what AI thinks my brand is known for?

Accepted Answer

Yes, on Starter, Growth and Pro, through Attributes. A weekly study, which owners and editors can also run on demand at most once a day, asks each chat engine on your plan, and ChatGPT (not Google AI Overviews, AI Mode or Gemini), what your brand and each active tracked competitor are best known for, and scores each quality by how early the engines list it. It then asks, without naming your brand, which brands in the market you name (for example “book-writing software for authors”) are best known for up to ten attributes you choose and the strongest qualities it found, so you can see whether engines that describe you as, say, easy to use also name you first for ease of use, and which brand leads, tracked or not. Each score is built from quotes located in saved engine answers, and the grouping of similar wording into one attribute is an AI model’s judgement. Studies are separate from your tracked prompts and change no visibility metric; until an owner or editor sets the market, only the first question is asked.

Question 15

Can DiscoveredBy check whether AI engines get our facts right?

Accepted Answer

Yes, on Starter, Growth and Pro. In Fact check, owners and editors keep up to 50 approved facts about their own brand, such as pricing, founding, availability and policies. A fact typed in is active at once; a draft suggested from the brand’s own site is checked only after someone approves it. A weekly fact study asks each chat engine on the plan, and ChatGPT (not Google AI Overviews, AI Mode or Gemini), one fixed question per topic you have facts for (except Other), written without your fact text so it does not lead the engine, and the answers to up to 10 tracked prompts you choose are checked as they arrive, with a limit of 200 checked answers per project a day. An AI model judges whether each answer supports or contradicts each fact; our code keeps a finding only when its quote is found in the answer and its fact is one of yours, and links the sources the engine cited for that sentence. Contradictions wait in a review queue, where your team confirms them, marks that the answer actually agrees with the fact, marks your fact as out of date (then edits or archives it) or sets them aside, and a contradicted fact raises an alert, at most once every 7 days per fact. Accuracy per fact is supported answers over answers that addressed it, shown as not addressed rather than 0% when none did, alongside each engine’s result in the latest study and what changed since the previous one. The verdict is an AI judgement to review, not a ruling; claims that match none of your facts, and facts about competitors, are not checked; and nothing is sent to the engines to correct them. Findings are also available as an export, through the customer API and through MCP.

Question 16

Can my agency read project data through API or MCP?

Accepted Answer

Yes. Starter, Growth and Pro include a read-only customer API for current prompts, completed collected answers, saved citations, daily metrics, brand reasons, objections, attributes, fact checks, weekly report digests and Explorer queries. Project owners and editors can issue keys for one client project, choose separate prompt, answer, citation, metric, reason, objection, attribute, fact check, report and Explorer read permissions, set an expiry, and revoke keys. Full saved answer text is available in bounded chunks. Citation records include source URLs, positions, bounded snippets and answer provenance. Brand reasons reads also require the owner’s Brand reasons entitlement, objection reads the owner’s Objections entitlement, attribute reads the owner’s Attributes entitlement, and fact check reads the owner’s Fact check entitlement. Weekly digests include the report’s stored metric readings, each with its definition (five for a current report, four for an older one), a bounded summary, activity counts and recommendation titles; report reads also require weekly-report access. Existing keys keep their original permissions. Each key allows 60 authenticated requests per minute, and every request checks current project, creator and plan eligibility. Daily metrics include collection counts, brand visibility and own-domain citation rate, with denominators and recorded collection provenance. Explorer reads, with the opt-in Explorer permission, compute any Explorer query or saved view on request and also count against the project’s shared Explorer budget of 120 queries a minute. A read-only MCP connector exposes the same datasets and permissions to Streamable HTTP clients that support a custom Authorization header. API and MCP share the key’s rate budget. OAuth-only clients, writes and a refreshable BI connector are not supported.

Question 17

Can I track AI visibility city by city?

Accepted Answer

Yes, on every plan. Add one or more cities to a tracked prompt, chosen from a list of 344 major cities across the 18 countries a project can target; to add a city that is not listed, contact support. A city is a location of its own, so it does not need its country selected, and each city, audience and language combination uses one prompt slot, the same as a country. The Local view compares each city with the same country country-wide over 7, 28 or 90 days: brand visibility, citation rate, share of voice and brand position, the gap between them, and the prompts where a city differs most. City is also an Explorer dimension. How the location reaches an engine depends on the engine: Claude, Perplexity and Grok receive it in a native location field and in the prompt text, Gemini API in the prompt text only, Google AI Overviews, AI Mode and Gemini as the city’s coordinates in the request to our licensed data provider, and ChatGPT as the country only, so a city target does not run on it. Each answer records which method was used. A target city is information passed to the engine, not the location of a real person using a consumer app. CSV and Excel import and accepting a suggested prompt add country-wide targets only.

Question 18

Can I build my own charts and dashboards?

Accepted Answer

Yes, on every plan. The Explorer combines up to five of twenty-two metrics (answers collected and analysed, brand visibility, brand position, share of voice, citation rate, citation share, answers citing sources, retrieval coverage, positive and negative sentiment, the citation rate of any cited domain, for the Google engines, AI answer shown rate, organic first page rate and organic position, and seven answer features: answers using web search and answers with shopping, local businesses, ads, video, images or tables) with up to two breakdowns and up to eight filters across engine, model, country, city, persona, language, intent, buying stage, theme, branding, tag, prompt, brand family and domain (the last two only with the metrics they apply to). Choose 7, 28 or 90 days ending yesterday or a custom window of up to 366 days, by day or by week, optionally compared with the previous period or the same weekdays 52 weeks (364 days) earlier and smoothed over seven days, and draw it as a line, bar, matrix, table or single-number tile. Missing data shows as no value, never zero, and small samples are marked provisional. Every project member can build and read queries; owners and editors can save up to 100 views per project and arrange them on up to 20 dashboards of up to 12 views each. With the project owner’s data-export access, any member can export a result as CSV or JSON. With weekly-report access, owners and editors can share a saved view or dashboard as a frozen snapshot link that expires after 1, 7 or 30 days, can be revoked, and never updates; anyone holding it can read it without signing in. There are no live public links, pie charts or BI connector.

Question 19

Can I ask questions about my data in plain words?

Accepted Answer

Yes, on Starter, Growth and Pro, through Analyst. Ask a question about the project’s tracked AI answers; Analyst plans up to four Explorer queries, runs them and writes a short answer with the charts and tables it used underneath. The model writes only the words: every number and date range is filled in from those results, and so is a row name wherever the model uses a row placeholder. The model still chooses which row and metric each number comes from, so each number shows its row and metric; the words around a number are not checked. An answer that types a number of its own is sent back for one rewrite; if the rewrite fails too, you get the results without a summary. Missing data reads "no data", never zero. Follow-up questions build on the conversation. Any project member can ask, one question at a time, up to 50 questions per project per UTC day. Conversations are private to the member who asked; owners and editors share a result by saving it as an Explorer view. Analyst reads only the Explorer’s metrics: it does not search answer text, cannot change settings or prompts, and is not available through the customer API or MCP.

Question 20

Can I share a report with a client who does not have an account?

Accepted Answer

Yes. With weekly-report access, project owners and editors can preview a fixed client-branded digest and create a link that expires after 1, 7 or 30 days. Before sharing, you can opt into up to five saved citation examples from that report’s week, with URLs, bounded snippets and collection details. The preview shows the sample and available record count; the examples are frozen into the new link and do not recalculate its metrics. Save up to 20 project report templates with selectable sections and a project default; preview the selected template before sharing. Existing links stay unchanged. Clients can read it without signing in. Anyone holding the link can forward it, and you can revoke it to stop future requests. Existing copies cannot be erased. Owners and editors can inspect the saved contents of issued links, including expired or revoked ones. The digest uses the client name and accent colour, retains DiscoveredBy attribution, and excludes raw answers, prompt lists, analytics revenue and team details. Individual links contain one project. Client reports also offers combined agency links for 2–10 projects in one completed week, using each project’s default template or standard digest. Preview every client before sharing. Current editing access and report eligibility are required for all included projects; losing either for one client makes the whole combined link unavailable. Metrics remain separate.

Question 21

Can each client project organize sources differently?

Accepted Answer

Yes. Source types separates source categories, page formats and ownership. It shows conservative rule suggestions with evidence from cited URLs and saved metadata. Project owners and editors can override a page or apply a domain category within that project. Other client projects keep their own labels, even when they cite the same URL. Unsupported or conflicting rule evidence stays unclassified; labels do not establish source quality.

Question 22

Can I check whether my robots.txt blocks AI crawlers?

Accepted Answer

Yes. On Starter, Growth and Pro, project owners and editors can check a URL on their project domain against five AI crawler policies. Robots diagnostics shows the matching rule and line number, separates search from training crawlers, and reports failed fetches as unknown. The policy-only quick check is unsaved. Saved page diagnostics adds HTTP, HTML structure, readability, response-time and JavaScript-render checks as DiscoveredByBot (single lab heuristics), dated evidence, retained history and a browser render compared against the raw HTML response, for selected URLs. Rechecks distinguish open findings from observed resolutions; unknown results do not establish resolution. Saved results link from the page issue queue and remain readable by current project members. Neither workflow establishes actual third-party bot access, visits or citations.

Question 23

Does DiscoveredBy crawl my site?

Accepted Answer

Only when someone asks it to, and that can be anyone: your project’s owners and editors can audit your project’s own site, and any visitor can run a free audit of any public site, yours included, at discoveredby.ai/site-audit without an account. Both use the same crawler and the same checks. A project’s site audit starts at your project’s URL and crawls up to the number of URLs you choose (100 by default). A free audit starts at the home page of the site the visitor enters and crawls up to 25 pages; a visitor who verifies an email address with a one-time code can upgrade it, once, to a new crawl of up to 100 pages. A site holds one free audit at a time. Its crawl data is kept while a crawl waits or runs, then deleted 48 hours after the crawl whose results it shows finishes, or at once if its first crawl saves no page; an upgrade crawl that saves no page is deleted at once and leaves the 25-page results to run out their own 48 hours. Only then can a new free audit of that site start. Until then, anyone who enters that site at discoveredby.ai/site-audit sees its results; the results page asks search engines not to index it and never shows who asked for the audit. Each crawl follows any redirect of its start page, then stays on that final host and follows its internal links, sitemaps and declared canonical and alternate URLs as DiscoveredByBot, and checks every page it finds against 238 technical SEO rules in 17 sections, comparing each page’s raw HTML with a browser render. robots.txt is read again before each URL: a URL it disallows for DiscoveredByBot is not fetched, a robots.txt that is simply missing (any HTTP 400 to 499 response other than 429) places no restriction, and a robots.txt that cannot be read (a server error, a network failure, a 429 rate limit, or a file that is HTML or not UTF-8 text) leaves that URL unfetched and marked failed. Each crawl also sends a bounded number of HEAD requests, at most 100, to check the status of resources and links its pages point to, including links to other sites, without crawling them, and downloads up to 50 social preview images, which may be hosted on other sites; before each of these it fetches robots.txt from that host, other sites included, and skips the request when DiscoveredByBot is disallowed or the file cannot be read. Every browser request stays on the public internet, on ports 80 and 443, with WebRTC’s non-proxied path disabled; a request to a private address is refused. Page scripts run during the render, and requests to common analytics and tag-manager services (for example Google Analytics, Google Tag Manager, Meta Pixel, Hotjar, Microsoft Clarity and Segment) are blocked, so a crawl does not add page views to those tools; first-party or self-hosted analytics, such as a server-side tag manager or analytics proxied through your own domain, are not blocked. Separately, Suggest from my site in Fact check reads your homepage and up to seven pages linked from it on the same site, once, when an owner or editor asks, at most once a day, to draft facts for your team to approve. Project owners and editors can start a crawl of up to 1,000 URLs when the project owner has the robots diagnostics feature (standard Starter, Growth and Pro); owners and editors can stop a running crawl regardless of that feature, and every current project member can read saved results. This does not measure real AI crawler traffic to your site, and DiscoveredBy does not edit your site.

Question 24

Which Reddit threads, YouTube channels and review sites do AI engines cite for us?

Accepted Answer

Social sources, on Starter, Growth and Pro. It reads accounts on Reddit, Quora, YouTube, TikTok, LinkedIn, X, Facebook, Trustpilot, G2, Capterra, GitHub and other recognised networks straight from the cited URL: no social page is fetched and no brand-name search runs inside one. Each account is labelled yours, from the profiles saved in Settings; a tracked competitor’s, from the profiles saved on that competitor’s page; or unlabelled. Pick a 7, 28 or 90 day window ending yesterday, pooled across engines or narrowed to one recorded channel. Every account gets reach (how many of your prompts cite it at least once), consistency (how often, across the prompts it reaches) and influence (reach multiplied by consistency), alongside a citation count and, separately, how its output rate (the share of completed answers citing it) changed against the equal prior period in percentage points, which is a change in how often it is cited, not a change in the citation count itself; an account seen only in the prior period still surfaces under the change, gains and losses sort orders. A separate list shows prompts where a competitor’s account was cited and none of your own were, capped at 50 with the true total shown. Parsing is bounded to the 5,000 most-cited social URLs per request; beyond that, counts are lower bounds and rates and changes are not reliable. A cited account is not a claim that it caused an answer, and a saved profile’s label is only as accurate as what your team entered.

Question 25

Can I track product lines or sub-brands separately?

Accepted Answer

Yes, on Starter, Growth and Pro. Settings → Brands lets you add a sub-brand under your own brand and under any active tracked competitor, one level deep; a sub-brand cannot itself have sub-brands. A brand family, a brand plus its sub-brands, still counts once per answer in share of voice, your own visibility and position, so an answer naming both Acme and its sub-brand Acme Pay is one answer for the Acme family, never two. Each active sub-brand also reports its own answers, its share of its family’s own answers, and its own sentiment: an indented row under its family on Sentiment trends always, and on Competitors’ “By brand mentioned” list whenever it was named at least once in the window; those figures overlap the family total rather than adding to it. Brand evidence inside cited pages, what an engine retrieved, and prompt classification all count a sub-brand’s names as its parent being present, never as a separate row. For a new answer, recognising a name as a brand or one of its aliases is the extraction model’s own judgment; adding an alias or a sub-brand instead moves any of the project’s own past untracked mentions that used exactly that name onto it, with no re-analysis of past answers. A brand can hold up to 10 active sub-brands, with a cap of 50 active sub-brands per project. There is no regex alias, and a sub-brand cannot be reparented or renamed; archive it and create a replacement instead, though an archived sub-brand’s own name and aliases stay reserved, so restoring it is the only way back to that exact name.

Question 26

What is Generative Engine Optimization (GEO)?

Accepted Answer

Generative Engine Optimization (GEO) is the practice of optimizing your digital presence so that AI answer engines (like ChatGPT, Gemini, Perplexity, Claude, Grok and Google’s AI Overviews) cite, reference, and recommend your brand in their AI-generated responses. Unlike traditional SEO which targets search engine result pages, GEO focuses on influencing how large language models select and present sources.

Question 27

How is GEO different from traditional SEO?

Accepted Answer

Traditional SEO optimizes for keyword rankings and backlinks on search engine result pages. GEO targets AI answer engines that synthesize information from multiple sources into direct answers. While SEO focuses on clickthrough rates, GEO focuses on citation rates, Share of Voice in AI responses, and entity-level authority that AI models use to select trusted sources.

Question 28

Which AI engines does DiscoveredBy track?

Accepted Answer

Eight engines: Gemini API, Perplexity, Claude and Grok, whose generated answers DiscoveredBy collects through the Gemini, Perplexity Sonar, Anthropic and xAI APIs, and Google AI Overviews, Google AI Mode, ChatGPT and Gemini, which a licensed data provider fetches from Google’s results, from chatgpt.com and from gemini.google.com for each prompt, along with the organic results beside an AI Overview. ChatGPT and Gemini are what chatgpt.com and gemini.google.com show a person who is not signed in, so ChatGPT is measured through its app only, and Gemini both through its API and as its app answers. Which engines run depends on your plan. Daily scheduled runs measure your tracked prompts; API outputs can differ from consumer apps, and when Google shows no AI answer for a search, that is reported as its own outcome, the shown rate.

Question 29

How does DiscoveredBy improve my AI visibility?

Accepted Answer

DiscoveredBy follows a structured process: discovery and intelligent crawling, daily AI search monitoring, deep results analysis, actionable recommendations, and continuous performance tracking. We identify gaps between what AI engines cite and what your site offers, then provide specific technical and content recommendations.

Question 30

What are DiscoveredBy’s AI agents?

Accepted Answer

DiscoveredBy runs a team of twelve specialized AI agents that work your visibility automatically: the Business Profiler, Prompt Discovery, Sentiment & Framing, Visibility Watchdog, Citation Gap Finder, Article Topic Planner, Article Writer, On-Page Optimization, the Search Plan, Revenue Attribution, the Growth Advisor, and the llms.txt Advisor. Together they turn monitoring into clear, prioritized actions. See the Agents page for what each one does.

Question 31

Will On-Page Optimization invent changes or edit my site automatically?

Accepted Answer

It does not edit your site automatically. Each retained change has at least one quote matched to saved source text, but that check does not verify every claim in the proposed draft. Review the facts and draft before using it; regulated topics carry additional human-review flags. You decide what to implement in your CMS. After you mark a change applied, your later citation rate on its prompts can be compared with the recorded baseline, without proving that the edit caused the change.

Question 32

How do I choose which pages to fix first?

Accepted Answer

The Page issue queue groups known owned pages from prompt mappings, saved citation gaps, optimization opportunities, saved page diagnostics and your newest completed site audit. Only your project’s newest completed site audit feeds the queue; a running or failed audit never hides the previous completed one. Sort by actionable issue count, known prompts or URL, filter by work status, and open linked evidence, drafts and recorded outcomes. A gap with a linked optimization for the same page appears once. Every project member can read saved work on every plan; existing action permissions still apply. The inventory reads up to 2,000 newest records per source and warns when limited. It is not a complete site crawl, technical audit or predicted-lift score, and no saved issues does not mean a page is healthy.

Question 33

Can I hand an optimization recommendation to my implementation team?

Accepted Answer

Yes. Export task packet on Optimizations downloads a Markdown file, and Download PDF task packet provides a paginated PDF, with the target page, current recorded status, saved draft formats, review flags, source quotes and an unchecked handoff checklist. Owners, editors and viewers can download using the project owner’s data-export entitlement, normally available on Starter and above. One source packet covers one recommendation, up to 100 changes and 1 MiB; PDF output also has a 250-page/8 MiB limit. Downloads are recorded in Activity. Owners and editors can separately approve up to 25 reviewed open drafts for handoff. Changed or ineligible drafts reject the whole batch; approval never marks edits applied. It is a saved copy, not a live task or approval: exporting does not apply changes, publish to a CMS or send anything to an external task tool. Review the current recommendation before implementing; stale and dismissed drafts are for reference only.

Question 34

Can DiscoveredBy publish to WordPress?

Accepted Answer

DiscoveredBy sends a finished article into a connected WordPress site as a new draft post; it never publishes, edits or deletes anything there. This applies to articles only, not page-level optimization recommendations, which stay Markdown or PDF task packets. The project owner connects a WordPress account with a WordPress Application Password; only the owner can connect, replace or disconnect it, and owners and editors can send a draft once connected. Any current project member can read the connection status and its send history. Sending is available on Starter, Growth and Pro, read from the project owner’s plan; Free and Trial are not included. See the WordPress guide for setup, error messages and duplicate-send handling.

Question 35

Which plans include which agents?

Accepted Answer

The Business Profiler and Sentiment & Framing run on every plan, so every project, including the free tier, starts with a suggested set of prompts to track and sentiment scoring from day one. Starter adds ongoing Prompt Discovery (the agent that keeps refreshing your list with new prompts to track) plus the Citation Gap Finder, the Article Topic Planner, the Article Writer, and On-Page Optimization (up to 5 page fixes a week). The Search Plan and GA4 Revenue Attribution are available on every plan, so you can connect Search Console and GA4 from day one. The Visibility Watchdog raises alerts in the app on every plan; on every paid plan it also emails the project owner an alert digest. The Growth Advisor (a narrative growth briefing, weekly or on request) and the llms.txt Advisor (which audits your llms.txt against the spec and writes you a compliant one) are included on every paid plan. Growth scales On-Page Optimization to the top 20 pages refreshed weekly, and Pro scales it across the top 80 pages weekly and unlocks the full engine coverage and higher limits.

Question 36

How does DiscoveredBy keep my data secure?

Accepted Answer

Role-based access control is enforced on every endpoint with strict tenant isolation, so you can only ever access your own projects. Traffic is encrypted in transit, session cookies are http-only and secure, database queries use bound parameters, and secrets and personal data are kept out of logs. See the Security page for details.

Question 37

Do you sell or share my data?

Accepted Answer

No. We never sell your data and we don’t share it between accounts. We collect only what’s needed to monitor your AI visibility, and your Google Search Console and Analytics connections use least-privilege, read-only scopes.

Question 38

When should I start optimizing for AI search?

Accepted Answer

Now. Billions of queries are already being answered by AI engines instead of traditional search. Early movers who optimize their content for AI citations will build compounding authority as these platforms grow. The longer you wait, the more ground competitors gain in AI-generated recommendations.

Question 39

Is DiscoveredBy suitable for small businesses?

Accepted Answer

Yes. Any business that relies on being found online benefits from AI search visibility. DiscoveredBy is free to get started. Whether you're a startup or enterprise, understanding how AI engines perceive your brand is essential as consumer behavior shifts toward AI-powered search.

Question 40

Do you offer a managed service?

Accepted Answer

Yes. Enterprise is a managed engagement from $1,000 a month with a 3-month minimum. It includes the platform with limits sized to your business, an onboarding audit in the first 2 weeks, 4–8 human-edited articles a month and a monthly strategy call with our co-founder, Kamal Karna. Digital PR, managed SEO, social media and custom integrations are quoted by scope. See discoveredby.ai/enterprise.

Frequently asked questions.

Start monitoring your AI visibility.