Lantern MCP connector

Get a link that connects Slack to your account. Someone has to open the link and approve it in Slack; the assistant cannot finish this step. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.connect_slack

End a session in which an assistant is working for this brand. The session stops at once and cannot be restarted.

Name your assistant sees: agent.terminate_delegated_session

Get a summary of your Lantern account: its name, plan, and how many members and brands it has.

Name your assistant sees: account-manager.get_account

Get your account's display name, logos and email sender name, as they appear on reports and emails.

Name your assistant sees: account-manager.get_account_branding

Get your plan, subscription status and next renewal date. Agency accounts also get their plan details and credit balance. A balance of null means it could not be read, not that it is zero.

Name your assistant sees: account-manager.get_billing_summary

See how many credits you have used this billing period, how many are included, whether overage is on, and a breakdown by brand and by team member. Projected usage for the rest of the period is not included; it is on Settings, Plan & usage. Check this before any action that uses credits. If this is not available, send the user to Settings, Plan & usage.

Name your assistant sees: account-manager.get_credit_usage

See your notification settings: email on or off, how often you get a digest, which topics you get, muted brands and quiet hours.

Name your assistant sees: account-manager.get_notifications

See the limits on your plan: how many brands, store connections and tracked prompts it allows. Check this before adding any of them.

Name your assistant sees: account-manager.get_plan_limits

See how many visits and conversions reached the brand's website over the last 30 days from AI assistants such as ChatGPT, Perplexity, Claude and Gemini, in total and for each referring site. The brand needs Google Analytics connected in Lantern.

Name your assistant sees: findings-interpreter.get_ai_traffic_insights

See how shoppers moved through the site, as the Attribution page shows it: the most common paths shoppers from AI assistants took, their most recent journeys, and the patterns Lantern found. Takes days or a start and end date.

Name your assistant sees: findings-interpreter.get_attribution_journeys

See AI traffic as the Attribution Traffic view shows it: visits by AI assistant over the window, their weekly trend over the eight weeks ending with the window, the split between AI and other traffic, AI agents fetching the brand's pages, and AI visits and sales by place when available. Takes days or a start and end date, and an optional assistant.

Name your assistant sees: findings-interpreter.get_attribution_traffic

Get the Attribution page overview for the brand: the revenue from shoppers who arrived from AI assistants, the recent orders they placed, and visits by AI assistant, over the page's window. Takes days (7, 30 or 90; the default is 30, as on the Attribution page) or a start and end date. Reports installed: false when the brand has no confirmed tracking pixel yet, and how fresh the results are. Includes a Google AI estimate when available, kept in Search and separate from AI visits.

Name your assistant sees: findings-interpreter.get_attribution_overview

Check whether Google Analytics is connected, and list the properties Lantern reads for the active brand.

Name your assistant sees: account-manager.get_ai_traffic_connections

List the changes the brand registered on its site and the measured effect of each on AI shopper behaviour, as the Attribution page shows them.

Name your assistant sees: findings-interpreter.get_attribution_changes

Give Lantern writing that shows how the brand sounds: pasted text, an interview transcript, or the address of one page, such as a style guide or About page. A page is read in the background and is ready in a minute or two; if it cannot be read, the source is still saved as failed, so do not add it again: use retry_voice_source to try again. To upload a file, use the brand voice settings in Lantern. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.add_voice_source

Find the pages on the brand's own website that best show how it writes (its story, collections, journal posts and a few products) and add them as brand voice sources. Runs in the background and takes a few minutes; it adds pages once per brand. To read a page that failed, use retry_voice_source; to add one specific page, use add_voice_source.

Name your assistant sees: account-manager.seed_website_sources

Answer the current brand voice interview question, point to a source you already added, or both. The result is the next question, or says the interview is done.

Name your assistant sees: account-manager.reply_voice_interview

Build a brand voice from the sources added so far. It runs in the background for a few minutes and ends as a proposed version for you to review; it never goes live by itself. Your credit balance is checked first. This does not take credits from it, but a voice that finishes building counts 5 credits toward the brand's monthly agent credit cap, if one is set. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.synthesize_brand_voice

Replace the live brand voice with a full voice document you supply. Lantern checks it first; if anything is wrong, nothing changes and the result points to the exact problem. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.update_brand_voice

End the brand voice interview and save everything answered so far as a source, even if some questions were skipped.

Name your assistant sees: account-manager.finish_voice_interview

Get the brand's current voice (tone, audience, phrases to avoid, key messages) and any new version waiting for review.

Name your assistant sees: account-manager.get_brand_voice

List the material the brand voice is built from (text, web pages, interviews, uploads) and whether each one has been processed. websiteDiscovery is running while Lantern is still finding pages on the brand's site; don't ask the user for pages then.

Name your assistant sees: account-manager.list_voice_sources

See what Lantern can do through a store platform's connection, such as inviting staff or syncing products. staffInvitations is false when staff invitations are not turned on for your account, even on a platform that supports them; staffInvitationsNote says why. Something missing from the list is unknown, not unavailable. The Lantern pixel works on every platform; use get_pixel_install_snippet for it.

Name your assistant sees: account-manager.inspect_platform_capabilities

Add a new brand to your account, with its name and at least one website. Your plan sets how many brands you can have; if you are at the limit, the result says so.

Name your assistant sees: account-manager.create_brand

Match a description like "my UK Shopify" to one of the connected stores. Returns the match, or a short list to choose from when more than one fits.

Name your assistant sees: account-manager.resolve_store_reference

Get one brand's name, short name and websites. Leave brandId out to use the brand selected for this connection.

Name your assistant sees: account-manager.get_brand

See who owns the brand and pays for it, whether tracking is active, whether a handoff to a client is waiting for an answer, and which people from outside the owning account have admin access. Handing a brand to a client is done from the brand's settings page, not here.

Name your assistant sees: account-manager.get_brand_ownership

Check whether a website you claimed is verified, and get the DNS record that proves you own it. Use the domain exactly as list_account_domains shows it. Only account owners and admins can see it, because the record is a secret to the claim.

Name your assistant sees: account-manager.get_domain_status

See how far the brand's setup has come: whether the first scans have finished, which setup steps are done, and anything blocking the rest.

Name your assistant sees: account-manager.get_onboarding_status

Show which theme is live on each connected store, when it was published, and its status. Pass store_id to check one store.

Name your assistant sees: account-manager.get_theme_status

Explains one cited page (pass url) or one cited source site (pass domain) for a brand: how often it is cited in answers to the brand's prompts, on which AI engines, for which prompts, and what the page says about the brand. Reads the same data as the citation detail page. Covers the last 7, 30 or 90 days (default 30) ending on the brand's last finished day of results, or a custom range with start and end (YYYY-MM-DD). Pass exactly one of url or domain. When you quote a number, say which window it covers. To show the page, call navigate_dashboard with tab "citations", citation_url or citation_source, and brand_id set to this result's brandId so the page opens under the brand you read; the page opens in the window the user already has selected, which can differ from the window you quoted.

Name your assistant sees: findings-interpreter.get_citation_detail

How citations are moving for the brand over the last 7, 30 or 90 days (default 30), ending on the brand's last finished day of results, or the days ending on end_date, the same window the Citations page shows: the top cited sites and pages with the window total and the change from the first half of the window to the second half, plus the brand's own pages ranked by citations (ownPages, always across every prompt). Use get_citations for the full list of cited pages. get_citation_trend_by_channel splits it by AI assistant.

Name your assistant sees: findings-interpreter.get_citation_trend

How citations are moving for each AI assistant that answered the brand's tracked prompts in the window: per assistant, the top cited sites and pages with the window total and the change from the first half of the window to the second half, plus the brand's own pages cited by that assistant (across every prompt). An assistant that did not answer in the window is not listed. Covers the last 7, 30 or 90 days (default 30) ending on the brand's last finished day of results, or the days ending on end_date.

Name your assistant sees: findings-interpreter.get_citation_trend_by_channel

List the pages AI assistants cited when answering the brand's tracked prompts, the same list the Citations page shows: each cited URL with its site, title, how many times it was cited and when it was last seen, its share, and whether it is the brand's own page, a competitor, social media or an article. Each row's share is of the citations this call matched (the window, search and page_class together), not of every citation in the window. Do not add up these rows under page_class to get a class's share of all citations: a page is classed by the most common of its own, competitor and social citations, or as an article when it has none of those, while a share counts every citation by its own type. Covers the last 7, 30 or 90 days (default 30) ending on the brand's last finished day of results, or the days ending on end_date. Returns up to 50 pages at a time; to get the next page, call again with next_offset as offset and the same search and page_class. get_citations_by_channel splits the same list by AI assistant, without page_class. There is no file download; when asked to export, give the list here and point to the Citations page for the full table.

Name your assistant sees: findings-interpreter.get_citations

The pages each AI assistant cited when answering the brand's tracked prompts, one entry per assistant that answered in the window (ChatGPT, ChatGPT (API), Perplexity, Google AI Overviews and others). Each entry has its own totals and its most cited pages, with each page's share of that assistant's citations among pages matching search, when one is given. An assistant that answered but cited nothing shows zeros; one that did not answer in the window is not listed. Covers the last 7, 30 or 90 days (default 30) ending on the brand's last finished day of results, or the days ending on end_date. To see more pages for one assistant, send its channel with offset and the entry's next_offset.

Name your assistant sees: findings-interpreter.get_citations_by_channel

Start tracking a competitor for the brand. A brand can track up to three. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.add_competitor

Add a starting list of competitors during setup, usually the ones AI answers already mention alongside the brand. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.seed_competitors

Start an AI readiness analysis of a competitor's website and wait for it to finish, with progress updates along the way. Pass the same idempotency_key again within a day to get the first run back.

Name your assistant sees: monitor.analyze_competitor

See how often AI assistants cite each competitor compared with the brand, on the prompts the brand tracks: citation counts, share of voice, typical position in the answer (lower is better), and which assistants cite them. This is relative to the brand's own prompts, not a market share.

Name your assistant sees: findings-interpreter.compare_brands

Confirm a competitor Lantern suggested, so it is tracked as one you chose. If you removed it earlier, add it again with add_competitor instead.

Name your assistant sees: account-manager.verify_competitor

List the competitors tracked for the brand. Only confirmed competitors are included unless you set includePending. Each competitor's id is what analyze_competitor takes.

Name your assistant sees: account-manager.get_competitors

Turn down a competitor Lantern suggested, so it stops appearing for this brand. Rejections also help Lantern suggest better competitors.

Name your assistant sees: account-manager.reject_competitor

Stop tracking a competitor for the brand. You can add it back later with add_competitor.

Name your assistant sees: account-manager.remove_competitor

Approve a draft that is waiting for review, so it is ready to publish. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: content-studio.approve_draft

Set a draft aside for good, so it can no longer be approved or published. Only drafts that are waiting for review or approved can be discarded. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: content-studio.discard_draft

Hide a content idea from list_content_suggestions for a while. It takes a content idea's id, not an opportunity's; to skip an opportunity, use dismiss_opportunity.

Name your assistant sees: content-studio.dismiss_content_suggestion

Write a piece of content for the brand: an expert guide, product story, FAQ, help article, community reply or pitchable article. It is written in the background over a few minutes and lands in your drafts for review; it is never published by itself. To write for a content opportunity, pass its opportunityId. The result has a link to the draft. Your credit balance is checked first. This does not take credits from it, but each draft that finishes counts 3 credits toward the brand's monthly agent credit cap, if one is set. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: content-studio.draft_content

Fetch and save one source for a Content Studio draft, or refresh its saved text. Supply the selected source ID, URL, current kind and both current revisions. Preserves other sources and proposals, resets affected decisions, and records blocked or unavailable sources without retaining old text. Never changes the document.

Name your assistant sees: content-studio.refresh_draft_research_source

Get a draft's full content, what it was based on, and where it will be published.

Name your assistant sees: content-studio.get_draft

Get a saved image and its source records for this brand.

Name your assistant sees: content-studio.get_content_image

Save an image from a web address to this brand’s private library. Importing does not grant permission to publish it.

Name your assistant sees: content-studio.import_content_image

Get one opportunity as the Opportunities page shows it: its summary, what buyers ask and what to do, and for a visibility gap, the question buyers ask, what AI answers say today and the sources they used. Takes an opportunity id and its kind from list_content_opportunities.

Name your assistant sees: content-studio.get_opportunity

Get before and after images of a page for one suggested fix. The links expire after about 15 minutes, so call again for fresh ones. An empty list means no mockups have been made yet.

Name your assistant sees: findings-interpreter.get_mockup

Get everything about one suggested fix: what kind it is, the exact change, when it was created or applied, and whether it was verified. Read this before applying or explaining a fix. When present, pass the returned mockupRecommendationId to get_mockup to see before and after images.

Name your assistant sees: findings-interpreter.get_action_metadata

See how much a suggested fix is expected to help, how confident Lantern is in that, and how large a change it makes, so you can weigh it before applying.

Name your assistant sees: findings-interpreter.get_action_impact

List the brand's open opportunities as the Opportunities page shows them: things to do to show up in more AI answers, most important first, each with its summary, what buyers ask and what to do, plus the drafts already waiting in Content Studio. Takes an optional kind. Says when the list is still being prepared.

Name your assistant sees: content-studio.list_content_opportunities

List up to 10 suggested fixes waiting for your review, newest first. For what to work on next, use get_overview_attention and list_content_opportunities.

Name your assistant sees: findings-interpreter.get_action_queue

Mark a content opportunity done after completing it. Repeating this action succeeds without changing the original completion time.

Name your assistant sees: content-studio.mark_opportunity_done

Rate a content opportunity as useful, not relevant, or wrong. Not relevant hides an open item for a while; wrong prevents it from returning on re-detection. An optional note explains your rating. Completed items keep their completion.

Name your assistant sees: content-studio.rate_opportunity

Get one persona by its id or exact name, including its products, pages, questions and stored need evidence. Evidence includes family labels and price ranges in currency minor units, whether one price range holds, top product pairs with joint and expected counts of visits, one per browser (its first visit in the window that viewed a product and opened no more than 20 different pages; see pairsNote), which are not comparable with core share counts, whether pairs held in both halves, whether orders are comparable, guide reading, the measured statement and core share counts. When present, area describes where the audience's located visits come from: the display grain, whether locations are approximate, and up to three named places. The separate income comparison states whether the median household income of the areas all of the audience's located visits come from is higher or lower than that of the areas the store's other located visits come from, with its Census ACS source. It covers all of those areas, not just the named places, so never say the named places have higher or lower incomes. Area covers at most the last 30 days up to the data-through date, starting no earlier than the store's first visit with a known county; window gives its first and last day, so name those dates when describing it. It describes places visits come from: metro or micropolitan areas, or states. Area figures count each browser once however many visits it made. Area is absent when there are too few located browsers, when the located rates of the audience and the store's other visits differ too much, or when it was not computed; absence means unknown, never none. When present, areaProfile describes where the audience's visits come from over its own window of at most 30 days: the share of browsers placed in a country, up to five world regions, countries and US states by share (its bases field says what each share is a share of), and averages of Census figures for the US counties those visits come from. Those figures describe counties, never a visitor and never the audience, they lean toward cities and suburbs, and they must not be compared with national figures. Do not rank audiences by these figures, and do not combine them with area, which is computed on a different base. areaProfile is absent when it is not enabled, was not computed, or has too few placed browsers; absence means unknown, never none. Not-computed evidence carries only status, window start, data-through date and definition version; every other field is unknown, never zero or none. namedBy null means channel appearances were not computed. When a figure is held back, give the reason exactly as returned; ordersAreLowerBound means the real order count may be higher. The Personas page in Lantern shows the same figures.

Name your assistant sees: findings-interpreter.get_audience

List the brand's personas, the groups of shoppers who behave alike, with the visits and orders recorded for each. When a figure is held back, say it is held back, never zero, and give the reason exactly as returned. Revenue across mixed currencies is not available. The Personas page in Lantern shows the same figures.

Name your assistant sees: findings-interpreter.list_audiences

Get the Products page summary for the brand: how many of its products AI shopping answers recommended, how many tracked prompts named its products, what share of the links on its products sent shoppers to its own site, which AI engines named its products, which competitors' products appeared most, and which stores shoppers were sent to. Takes days (7, 30 or 90; the default is 30, as on the Products page). Other values clamp to 1 to 90. Also reports how fresh the results are. When dataThroughUnavailable is true, the freshness could not be read and the window runs through today; do not treat dataThrough as real.

Name your assistant sees: findings-interpreter.get_product_catalog_summary

See one product the way its detail on the Products page shows it: how many tracked prompts named it and how that compares with the previous window, the prompts and AI engines that named it, the competing products named beside it and how often it and they appeared on each day, the stores AI answers sent shoppers to and the prices they quoted, the share of its links that went to the brand's own site over time, the kinds of site its links went to compared with its rivals, and the latest price and rating an AI answer quoted for it. Pass the product_id or product_key from get_products. Takes days (7, 30 or 90; the default is 30, as on the Products page). Other values clamp to 1 to 90. Also reports how fresh the results are. When dataThroughUnavailable is true, the freshness could not be read and the window runs through today; do not treat dataThrough as real.

Name your assistant sees: findings-interpreter.get_product_detail

List the products on the brand's Products page, in the page's order. When the brand's catalog has been imported and matched, this is the catalog, and filter picks the same groups as the page's filters; otherwise it lists the products that appeared in AI shopping answers. Each product has a product_id or product_key to pass to get_product_detail. Takes days (7, 30 or 90; the default is 30, as on the Products page; other values clamp to 1 to 90), an optional search text, and a limit from 1 to 50 (the default is 20). Also reports how fresh the results are. When dataThroughUnavailable is true, the freshness could not be read and the window runs through today; do not treat dataThrough as real.

Name your assistant sees: findings-interpreter.get_products

Bring in the latest products from a connected Shopify store. Your plan limits how many products Lantern keeps for each store; products past that limit are archived. The result counts what was created, updated and archived.

Name your assistant sees: monitor.sync_store

Add a tracked prompt to one of the brand's collections. Adding one that is already there changes nothing. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.add_prompt_to_collection

Hide one of Lantern's prompt suggestions so it stops appearing.

Name your assistant sees: account-manager.dismiss_suggestion

Change a tracked prompt: its wording, whether it runs, whether it is a favorite, or whether it counts as a branded prompt. Give only what you want to change. New wording that matches another prompt the brand tracks is refused.

Name your assistant sees: prompts.edit_prompt

Estimate how many credits a brand's tracking will use each month, with the prompt count, schedule and AI assistants behind the number. Nothing is charged. Share this before turning tracking on.

Name your assistant sees: account-manager.estimate_visibility_credits

Read what AI assistants actually answered for one tracked prompt. With no dates, you get the latest answer from each assistant; with start_date or end_date, every answer in that window, newest first. Identify the prompt by prompt_id or by its wording. Each answer is marked "answered" or "no_answer". "no_answer" means the assistant gave no answer (for google_ai_overview, Google showed no AI Overview for that search). It is not a miss: its text is empty, mentioned and rank are null, and it counts toward no visibility score. mention_checked false on an answered response means Lantern has not checked that answer for the brand, so mentioned and rank are unknown, not false; Prompt detail shows the newest checked answer instead. When the brand has a finished day, answers stop at that day, as on Prompt detail. Also reports how fresh the results are. When dataThroughUnavailable is true, the freshness could not be read and answers run through today; do not treat dataThrough as real.

Name your assistant sees: findings-interpreter.get_prompt_responses

Ask Lantern for new prompt suggestions for the brand, researched the same way as its weekly suggestions. They appear in list_suggestions for you to track or dismiss. Your credit balance is checked first. This does not take credits from it, but each run that finds new suggestions counts 5 credits toward the brand's monthly agent credit cap, if one is set. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: account-manager.propose_suggestion

See how tracking is set up for the brand: whether it is on, how often it actually runs, when it runs next, and how many prompts it tracks.

Name your assistant sees: account-manager.get_brand_visibility_settings

Set a tracked prompt's category, collections or funnel stage; only what you give changes. Each label is saved separately, so if one fails, the result lists the labels that were saved.

Name your assistant sees: prompts.set_prompt_labels

Estimate how many credits a report will cost and whether your balance covers it. Nothing is charged.

Name your assistant sees: account-manager.estimate_report_credits

Check whether a report is still being made, is a draft, is published or failed. Published reports include their share and PDF links.

Name your assistant sees: account-manager.get_report_status

List the reports made for the brand.

Name your assistant sees: account-manager.list_brand_reports

Run a fresh analysis to find new suggested fixes and store tasks. Your credit balance is checked first; this does not take credits from it.

Name your assistant sees: site-editor.re_analyze_actions

Record that you made a suggested fix yourself, outside Lantern.

Name your assistant sees: site-editor.mark_recommendation_done

Put back what was live before a change Lantern published, using the snapshot taken at the time. This replaces what is live now. Depending on the brand's approval settings, the change may wait for someone to approve it in Lantern; the result then says it is proposed.

Name your assistant sees: site-editor.rollback_change

Start an AI readiness analysis of the brand's website and wait for it to finish, with progress updates along the way. Pass the same idempotency_key again within a day to get the first run back instead of starting another.

Name your assistant sees: monitor.analyze_domain

Get a summary of every brand you can see: each one's status, whether its first analysis has finished, how many prompts are running, and the last day its results run through, and whether that day came in complete. When the brands span several accounts, each one also names the account that owns it (ownerAccount) and whether it is in the account you are working in (inCurrentAccount); waitingForClientEmail is set when you have sent a client an invite for that brand and it has not been answered. When a brand's resultsUnavailable is true, its dataThrough and resultsState could not be read; do not describe them. Use this to compare brands.

Name your assistant sees: findings-interpreter.get_portfolio_overview

Get the brand's latest site analysis scores and analysis status in one call. detail "summary" (the default) answers most questions; "full" adds a page breakdown; "with_history" adds the last 12 weeks of scores; "with_per_query" adds results per prompt, including which competitors were cited. Only go past "summary" when asked for that detail. For what to work on next, use get_overview_attention and list_content_opportunities instead.

Name your assistant sees: findings-interpreter.get_brand_overview

See where the brand stands against the rivals it tracks, as the Brand Positioning page shows it: the brand and up to six of its tracked rivals in the page's order, whether each moved up, down or stayed steady and who is closing in, where the brand and the tracked rivals with enough data stand in words, the other brands also seen in answers, and each day's Presence over the window. Takes days (7, 30 or 90; the default is 30, as on the page) or a start and end date, and the page's filters.

Name your assistant sees: findings-interpreter.get_brand_positioning

See which words and phrases AI assistants use to describe the brand are rising and which are fading, and whether each one reads as positive, neutral, negative or mixed. Compares the recent half of the last 12 weeks with the earlier half.

Name your assistant sees: account-manager.get_descriptor_trends

How the three headline figures on the Overview moved, day by day: the visibility score (0 to 100; it rises when more answers mention the brand and when they place it higher), presence (the share of answers that mention the brand) and average position (where the brand first appears in answers that rank it; 1 is first). Returns one row per day tracking ran; a day with no answers is left out, never shown as zero. The summary gives the score on the latest day with answers and its change from the previous window, presence for the whole window with its change, and the average position once at least 8 answers rank the brand. Daily position has no minimum. Covers the last 7, 30 or 90 days (default 30) ending on the last finished day of results, or the days ending on end_date. Use prompt_ids to limit the figures to some tracked prompts. For the same figures per AI assistant, use get_brand_overview_trend_by_channel.

Name your assistant sees: findings-interpreter.get_brand_overview_trend

The same daily visibility score, presence and average position as get_brand_overview_trend, for each AI assistant that answered the tracked prompts in the window. Each assistant gets its own summary and its own rows, one per day it answered; an assistant that did not answer, or answered only branded prompts (which none of these figures count), is left out. The average position for a window is given once at least 8 answers from that assistant rank the brand; daily position has no minimum. Takes the same days, end_date and prompt_ids inputs as get_brand_overview_trend.

Name your assistant sees: findings-interpreter.get_brand_overview_trend_by_channel

See how the brand does by AI channel, category and collection, as the Brand Positioning page's segments show it: who leads each segment and where the brand leads or is behind. With a dimension and a segment id, see that segment's prompts and how the brand does in each. Takes days or a start and end date, and the page's filters.

Name your assistant sees: findings-interpreter.get_positioning_segments