How can we help?

One free complete AI search and report after you verify your email. No card required. It asks your customers' questions across the default AI sources and keeps every full answer, the sources each answer cites and the competitors it names, so you can read what AI says about your business and why.

No. You enter your website, the market you sell in and the questions your customers ask; IndexHalo suggests questions from your type of business and place, and you can edit or remove any of them. Findings are written in plain English with the answers behind them.

Every search asks ChatGPT, Gemini, Google AI Mode, Claude, Perplexity and Google AI Overviews by default, for one credit per question. Google organic results can be collected as their own search, for one credit per question, and are kept apart from AI answers. Claude and Perplexity answers come from their APIs with web search, which can differ from a signed-in conversation. Plan and top-up credits expire at the end of the billing month and do not roll over. Credits bought as one-time packs before monthly plans launched keep their balance and are shown separately.

Yes. Each saved question keeps its own country and language, and Copy to markets adds it to other countries without translating its wording. Nothing preselects a market: you choose each one.

As often as you run them. Professional includes 10 questions re-checked weekly; Business 30; Agency 150. No credits are used. Beyond those, schedule a saved question set daily, weekly or monthly. When you set the schedule, you approve a credit limit per run and per month; each scheduled run uses one credit per question and market.

No. It reads public pages and AI answers. Drafts and page changes stay in IndexHalo until a person approves publishing them.

Start with the actions: each one names the questions and answers it comes from. Once you have changed a page, Recheck now asks the same questions again with the same sources, countries and languages, so you compare matched answers rather than a different set.

Yes. Invite them as members of your workspace with a role; a viewer can read reports without changing anything. You can also send a read-only link to a fixed report, and revoke it at any time.

No. IndexHalo counts being named, being recommended and being cited separately, each with its own count and the number of answers it is out of.

The Overview flags answers that cite a site named like you or a competitor but not the website you saved. Confirm it is a different company and the affected questions in that search can be re-run once, free, with a dated correction notice on the original report.

IndexHalo asks your buyers' questions across ChatGPT, Gemini, Google AI Mode, Claude, Perplexity and Google AI Overviews, keeps every full answer and citation, and turns them into findings, actions and reports.

Every search asks ChatGPT, Gemini, Google AI Mode, Claude, Perplexity and Google AI Overviews by default, for one credit per question. Google organic results can be collected as their own search, for one credit per question, and are kept apart from AI answers. Claude and Perplexity answers come from their APIs with web search, which can differ from a signed-in conversation.

Yes. REST API v1 with scoped, revocable API keys and an OpenAPI 3.1 description at /api/v1/openapi.json. Create keys in Settings › Integrations. Keys can read data, create draft actions and save tracked questions (saving starts no search); no key can publish to a website.

Yes. A private MCP endpoint at https://www.indexhalo.com/mcp lets your own AI assistant read your IndexHalo data. Claude and ChatGPT connect with one URL and one approval, limited to the brands you choose; other assistants use a scoped API key. Its 10 tools are list_brands, get_results, list_questions, search_answers, get_actions, get_reports, what_changed, cited_for_competitors_not_you, list_markets and quote_search. Read-only: it can read your retained answers and reports; it cannot run searches or spend credits. It does not register your brand with any AI service.

Yes. Pull IndexHalo's CSV feeds into a Google Sheet with IMPORTDATA, then add that Sheet as a Looker Studio data source. A Looker Studio Community Connector is available on request as Apps Script source that you deploy once. Excel and Power BI read the same feeds with Get data › From Web.

Yes, on every plan. Reports export with your own name, logo and accent colour, and you can share a read-only client portal link that expires after 30 days by default, or up to a year.

Professional includes 10 questions re-checked weekly; Business 30; Agency 150. No credits are used. Beyond those, schedule a saved question set daily, weekly or monthly. When you set the schedule, you approve a credit limit per run and per month; each scheduled run uses one credit per question and market. Report emails and a weekly digest can be scheduled too.

The interface, reports and emails are in English. You can track questions in 13 countries and 17 question languages; each saved question keeps its own market and language, and every answer is kept word for word in its original language. Tone, recommendation, reputation and positioning rules read English, German, French, Spanish and Dutch answers. Spanish wording has been checked on test passages only, not yet on real answers. Answers in other languages show 'not classified (language)', never a zero. Buyer routing, Why-this-answer criteria, concern sources and fact disagreements read English answers only for now.

Upload a GA4 export to see AI referral visits by landing page. Search Console and Bing Webmaster exports can be imported too. Live GA4 and Search Console sync is built and available once connected.

No. IndexHalo reports counts with their denominators, for example 'Named in 31 of 50 answers', and keeps being named apart from being recommended. Change is shown only when the same questions were checked again under the same scope.

Monthly plans: Professional $49 for 100 credits, Business $129 for 300 credits and Agency $299 for 750 credits a month, plus tax. Every plan has every feature. Subscribers can add top-up packs at the same prices. Plan and top-up credits expire at the end of the billing month and do not roll over. Credits bought as one-time packs before monthly plans launched keep their balance and are shown separately.

One free complete AI search and report after you verify your email. No card required.

Yes. One free complete AI search and report after you verify your email. No card required. Your account saves the results so you can return to them later.

Yes. Accounts support password reset links when transactional email is configured and a private recovery-code route that does not depend on an external identity provider.

IndexHalo's monthly plans, in US dollars, are Professional $49 for 100 credits, Business $129 for 300 credits and Agency $299 for 750 credits. Every plan includes the same full package; only the monthly allowances change: credits, Ask your data questions, page-fix drafts and question generator runs. Subscribers can add top-up packs of 100 credits for $49, 300 credits for $129 or 750 credits for $299. One credit covers one buyer question, one market and language, one run, across the selected available sources, with its analysis and saved report. Checkout shows the final amount before purchase.

Yes. Every search asks ChatGPT, Gemini, Google AI Mode, Claude, Perplexity and Google AI Overviews by default, for one credit per question. Google organic results can be collected as their own search, for one credit per question, and are kept apart from AI answers. Claude and Perplexity answers come from their APIs with web search, which can differ from a signed-in conversation. A credit is consumed only after every selected source has usable evidence and the complete report is saved. Failed or incomplete searches release the entire reserved credit. Reopening or exporting saved reports is free.

One free complete AI search and report after you verify your email. No card required. A complete AI search measures the answers the AI sources actually return.

Yes. Web plans renew monthly. Unused monthly credits reset without rollover. Cancel renewal in Billing at any time; the subscription continues until the end of the paid period. Plan and top-up credits expire at the end of the billing month and do not roll over. Credits bought as one-time packs before monthly plans launched keep their balance and are shown separately. There are no automatic overage charges.

There are no automatic overage charges. New searches pause when no plan, top-up or earlier one-time pack credits remain. Subscribers can add a top-up pack or upgrade in Billing, or wait for the next paid renewal. Saved reports remain available to read and export.

Stored account and historical report data are handled separately under the privacy and deletion policies. A verified refund may adjust unused credits but does not silently rewrite completed historical evidence.

No. The free website audit is no longer offered, and audits saved earlier stay readable under Reports → Page audits. What is free today is one complete AI search with a newly verified account: one buyer question across the available AI sources, with the answers, cited sources and saved report. No payment card is required.

The signed-in workspace adds Chats with every full answer, Why this answer, Tone and identity, Brand shape, Reputation, Accuracy (fact review), Locations, Products, Ranking, Compared brands and share of voice, Gap analysis, Cited domains and pages, Discovery, Fanouts, Crawlability, Crawler logs, AI referrals, Actions with evidence and verification, Impact, Scheduled searches, Report exports, White-label exports and Client portal links, plus the integrations listed at /integrations/.

It is a signed-in view in Results › Chats › Fanouts. The searches and pages AI used for questions where you were missing, turned into briefs for the content to create or strengthen. Copy a brief as a task or download the plan; no new search is started.

Yes. Complete AI visibility reports export as PDF, HTML, Word, PowerPoint and Markdown, with audience and evidence options. Website audit reports support JSON, Markdown, page-inventory CSV and print/PDF. Analytics views also offer their own evidence downloads. Configured report branding carries into supported client exports, and website monitoring preserves dated snapshots for comparison.

Yes. Signed-in reports can import CSV exports from Google Search Console, Bing Webmaster Tools, and Google Analytics 4. IndexHalo maps measured impressions, clicks, CTR, position, users, conversions, and revenue to crawled pages and buyer questions, then ranks concrete opportunities. Upload a GA4 export to see AI referral visits by landing page. Search Console and Bing Webmaster exports can be imported too. Live GA4 and Search Console sync is built and available once connected.

It is a signed-in view in Agent analytics › Crawlers & referrals › AI referrals. Upload a GA4 export to see visits that arrive from AI assistants, by landing page; direct and unattributed visits are never counted as AI. Only rows whose supplied session source, source / medium, referrer or campaign explicitly identifies an AI answer product are classified, with sessions, engagement, conversions and revenue where the export includes them.

No. Direct, unattributed, organic, unknown and not-set rows are never redistributed across AI providers. The report intentionally understates influence when the source is unavailable rather than inventing a larger attribution number.

No. A referral is measured analytics evidence; a citation is a separate timestamped provider observation. IndexHalo can display them beside each other as an observable funnel, but it does not claim that one caused the other.

Include landing page, session source or session source / medium, sessions, engaged sessions, key events or conversions, and revenue. Full referrer, channel group and campaign are useful supporting dimensions. Add a dated window and use the same dimensions for later comparisons.

Only when a previous GA4 import exists and both imports contain dated windows of equivalent length within one day. Otherwise it shows the current measured values and withholds the comparison.

No. It shows revenue recorded on sessions explicitly identified with an AI answer-product source in the supplied export. It is not universal multi-touch attribution and does not prove that GEO work, a citation, or any single content change caused the conversion.

For each measurement source, the newest import is the current reporting window and the prior import is used for change comparison. The report labels these values as imported measurements and keeps them distinct from observed answers and citations.

Start with access failures and high-value pages, assign the recommended owner, implement the target condition, then recrawl and verify the stated test. Repeatable evidence matters more than any single number.

It is a signed-in view in Results › Ranking › Share of voice. Counts how often you and each compared brand are named with the denominator shown, for example 'Named in 31 of 50 answers', source by source. Being recommended and being cited are counted separately from being named, each with its own count and denominator.

Without the exact brands, questions, providers, timestamps and successful-response count, a percentage cannot be reproduced or compared. IndexHalo preserves that universe and excludes changed runs from movement rather than presenting them as progress.

Each rate is reported on its own as a count over its denominator, for example 'Named in 31 of 50 answers'; average numbered-list position is the mean position over the answers that number the brand, shown with how many there are. Nothing is weighted, averaged together or added into one score. Visibility: Answers mentioning this brand / AI answers with a known mention outcome. Share of voice: This brand's appearances / appearances of every configured brand, at most once per brand per answer. Not market share. Citation rate: Answers citing a configured domain of this brand / AI answers with known citation ownership. Average numbered-list position: Mean explicit numbered-list position, only where the answer actually numbers this brand. Lower is earlier; unranked mentions remain unknown. Ranked win rate: Answers where this brand has the earliest numbered position among tracked brands / answers with an unambiguous numbered ranking. Requires at least two configured brands; ties are unknown. Recommendation rate: Explicit positive recommendations / AI answers with a known recommendation outcome. Ambiguous language is excluded. An answer whose outcome could not be read stays out of the denominator and is shown as unknown, never as zero. Being named is not being recommended.

No. It describes only the configured brands, fixed questions, successfully verified responses and provider set. It is not prompt volume, total audience reach, revenue share, training inclusion, private model preference or general market share. IndexHalo does not combine mention, recommendation and citation counts into one index or score.

Competitors come from domains deliberately configured for the monitored website. IndexHalo uses exact domain matches in citations and conservative aliases derived from the configured domain or measured public title. Ambiguous aliases are rejected and the final brand universe remains visible.

Yes. IndexHalo preserves original Unicode questions and responses and applies disclosed recommendation and stance rules to English, German, French, Spanish and Dutch answers; answers in other languages show 'not classified (language)'. Local brand spelling, evidence, regulation, pricing, units and review still require the appropriate market owner.

No. The /100 page score was retired on 21 October 2026. Each page check reports pass, fail or unknown with the exact text it read; nothing is weighted, averaged or added up.

Direct findings can include response status, readable text, headings, authorship, dates, canonical metadata, structured data, quantitative claims, source links, crawler rules, redirects, and document extraction results.

A modelled estimate is a clearly labelled planning projection calculated from measured content, demand, cost, or conversion inputs. It is kept separate from archived answer observations and is never presented as a provider response, private ranking factor, or guaranteed outcome.

It is a timestamped provider response for a stored query, including the returned citation URLs, target-domain position, and competing domains. It proves what that response returned at that time and nothing more.

No independent GEO tool can reliably prove inclusion in undisclosed private training data. IndexHalo does not make that claim. It can document public crawl conditions and citations that appear in observable provider responses.

No. Search and answer systems change, personalise results, and evaluate many competing sources. IndexHalo identifies controllable gaps and verifies implementation, but it does not promise a future vendor decision.

Using the same declared question across compared sources improves comparability. Differences in answers, mentions, citations, and rank are interpreted only inside that retained question, source, locale, and time scope.

It is a signed-in view in Sources › Cited sources › Domains. Every domain and URL the answers cited, with how often and for which questions and AI sources. Citation trends in Sources › Cited sources › Citation trends shows which sources the answers cite, in how many answers, and when each was first and last seen.

No. IndexHalo keeps the exact returned URLs and domains with each answer and shows them in Sources › Cited sources. A returned source must be reviewed before a team relies on it.

It means producing genuinely useful, independently verifiable evidence—such as an original dataset, transparent method or named expert context—and making it available for legitimate editorial review. IndexHalo does not recommend buying, fabricating or implying endorsements or citations.

No. Repeated appearance is an observable pattern in a fixed test set, not proof of accuracy, trust, endorsement, traffic, backlinks or causal influence. Review the source, exact URL, claim, date, method and limitations before acting.

It is a signed-in view in Brand › Accuracy. Compares answers with the facts you approve, such as prices, features, hours and locations, and lists possible conflicts for a person to review. Claims that need review are collected in Brand › Reputation with the answers that made them.

No. Agreement shows that multiple verified responses expressed a similar statement in the fixed test. Providers may repeat the same stale or incomplete information. Teams must verify the statement against current primary evidence before relying on it.

No. Different values create a verification candidate. The values may use different years, markets, populations, units, scenarios or methods. IndexHalo preserves every exact statement and source URL so a research owner can resolve the scope.

Only when the provider returns durable inline attribution. Otherwise IndexHalo labels the URLs as response-level citations and does not claim that one source supports every answer sentence.

It is a signed-in view in Brand › Perception › Tone & identity. Labels how each answer speaks about your brand (positive, neutral, negative or mixed) from attributable passages in English, German, French, Spanish and Dutch answers, and shows the attributes answers connect to you. Every label points to a stored answer excerpt, question, AI source and time.

No. IndexHalo reports named-and-cited, named-without-citation, cited-without-name, and neither as separate states. Recommendation language is measured independently because a mention alone is not an endorsement.

No. Numeric, dated, absolute, comparative, or configured risk statements become verification candidates. A human compares the exact excerpt with current canonical evidence before deciding whether any correction or clarification is needed.

Yes. Exact configured aliases work across Unicode text, while disclosed recommendation, stance, and attribute rules read English, German, French, Spanish and Dutch answers; answers in other languages show 'not classified (language)'. The original provider excerpt remains visible so a fluent reviewer can validate every rule-based label.

Movement requires the same versioned buyer-question portfolio and provider set. Runs with changed denominators or no verified response text remain stored but are excluded from the trend with an explicit reason.

It is a signed-in view in Prompts › All prompts › Discovery. Lists questions and search phrases found in your retained answers that you do not track yet, ready to save as prompts. Setup also suggests buyer questions, and Search Console or Bing exports add search demand in Prompts › All prompts › Search demand; suggestions are not measured search demand.

No. Imported impressions measure exposure in the supplied search-platform period. IndexHalo never relabels them as model queries or total AI demand. Provider observations remain a separate evidence layer with their own exact denominators.

Save it as a prompt from Prompts › All prompts › Discovery. Its next search is quoted before it runs. Earlier checks stay stored, and change is shown only between matched repeat checks of the same questions.

Yes. Discovery reads the questions and search phrases in your retained answers. A Search Console or Bing export adds search demand beside your tracked questions; without one, demand is shown as not imported, never as zero.

No. Citation coverage and returned-source share describe the fixed questions, providers and timestamps in the monitored sample. IndexHalo publishes the denominator and does not present those observations as universal AI market share.

No. Citation coverage, position and the share of returned citations are reported separately, because a blended index hides which signal moved. Coverage and share each come with their count and denominator; position is a mean over the answers that cite you.

A provider error is unavailable evidence, not a negative citation result. IndexHalo reports attempted and verified checks separately and does not turn an incomplete response into 'not cited'.

Longitudinal movement requires the same versioned buyer-question portfolio and provider set. Runs with a different denominator or no verified evidence remain in history but are excluded from movement calculations with a stated reason.

Source convergence is the overlap between domains returned by different observed providers for the monitored questions. It helps identify evidence patterns shared across providers, but does not imply a private vendor relationship or prove why a source was selected.

Recommendations combine the observed gap, number and value of affected pages, expected signal impact, implementation effort, and confidence. Workspace actions include an owner, current state, target state, and verification test.

Yes. Whole-site crawling reads public crawler guidance, stays on the submitted domain, rejects private-network targets, limits page and response size, and records blocked or failed pages instead of hiding them.

It can read declared XML sitemaps and follow same-domain links from reachable pages. The report distinguishes discovered, analysed, redirected, blocked, duplicate, non-content, and failed URLs.

The current evidence crawler focuses on safely retrievable server responses and document structure. Heavily client-rendered content may expose less evidence than a server-rendered or progressively enhanced page, and the report records that limitation.

IndexHalo is designed for public material and deliberately rejects private-network destinations. It does not bypass authentication, paywalls, or access controls.

Yes. Signed-in reports accept common Apache or nginx logs, CSV, JSON, and NDJSON exports and can receive sanitized real-time crawler events through a rotating per-site secret. Recognized requests are mapped to the crawled page inventory, response errors, coverage gaps, and verifiable actions.

No. The crawler-evidence pipeline discards IP addresses and the raw log body before saving the normalized evidence. It retains only the request time, method, same-site URL or path, HTTP status, response-byte count, declared user agent, and crawler classification.

Not by itself. User-agent strings can be imitated. IndexHalo labels this as a request declaring the named crawler and does not present it as independently verified network identity, model training, citation, or downstream use.

The website accepts PDF, DOCX, TXT, Markdown, CSV, JSON, HTML, PNG, JPG, WEBP, and TIFF files up to the displayed upload limit.

Yes. When a PDF contains little or no extractable text, OCR processes its rendered pages and records the extraction path so the result is not confused with native document text.

No. OCR quality depends on resolution, contrast, language, layout, and scan quality. Important findings should be checked against the original document, especially when numbers or tables are involved.

The interface, reports and emails are in English. You can track questions in 13 countries and 17 question languages; each saved question keeps its own market and language, and every answer is kept word for word in its original language. Tone, recommendation, reputation and positioning rules read English, German, French, Spanish and Dutch answers. Spanish wording has been checked on test passages only, not yet on real answers. Answers in other languages show 'not classified (language)', never a zero. Buyer routing, Why-this-answer criteria, concern sources and fact disagreements read English answers only for now.

Original uploads are processed for the requested analysis and are not retained as a customer document library. Generated report data may be stored for signed-in report history according to the privacy policy.

Yes. Signed-in website reports convert ranked findings into a persistent action queue inside the Intelligence Hub. Teams can add an assignee, due date, implementation evidence, and workflow status while preserving the original measured finding, definition of done, affected pages, and verification rule.

It is a signed-in view in Optimize › Impact. Before-and-after views of the work you shipped, from matched checks; an association, not proof of cause. A change is compared only when the same questions were checked again with the same sources, market and model.

Yes. Choose the events in Settings › Integrations, for example a finished search, a paused schedule, a client that needs attention, a verified or regressed action, a recurring fact or reputation issue, or a watched answer that changed. They arrive in the in-app inbox, by email when you opt in, in Slack or Microsoft Teams, or as signed webhooks.

An action that cites answers is marked verified only when a later, exactly matched check (the same question, AI source, market and model) shows the change; until then it reads implemented and unverified. A later matched check can also mark it regressed. Verification is an association, not proof of cause.

Between matched repeat checks, IndexHalo shows which brands are named, recommended and cited, the cited sources that persist, arrive or fade, and the answers that changed. A single check shows what was found, never a trend.

Weekly is a useful default for active publishers. Daily suits fast-changing news, product, or launch environments; monthly may suit stable reference material. Run an extra crawl after migrations or major template changes.

No. IndexHalo does not sell submitted content. The privacy policy explains the operational data used to provide reports, accounts, security, analytics, and payments.

Google Analytics is configured with Consent Mode and advertising storage remains disabled. Region-appropriate analytics defaults apply, and every visitor can accept or reject optional analytics from the website privacy banner.

The labels stop direct findings, live observations, and estimates from being blended into a misleading claim. Professionals can see exactly what supports each conclusion.

The answer-experience catalogue is versioned, and every archived answer records the declared target and returned model identifier when available. Target changes affect new observations only; historical answers keep their original identity so comparisons remain interpretable.

Yes. Answer engine optimisation (AEO) and generative engine optimisation (GEO) describe the same underlying work — making content easy for AI systems to retrieve, verify and cite — with slightly different emphasis. Every IndexHalo check applies to both, and the same is true if your team calls the practice LLMO, AI SEO or generative search optimisation.

No. GEO builds on technical SEO, information architecture, quality content, authority, and user experience. It adds passage-level evidence analysis, citation observations, and stronger provenance and measurement requirements.

IndexHalo is AI visibility software. IndexHalo asks your buyers' questions across ChatGPT, Gemini, Google AI Mode, Claude, Perplexity and Google AI Overviews, keeps every full answer and citation, and turns them into findings, actions and reports.

IndexHalo is for agencies, SEO, content, PR and marketing teams that need to show clients and colleagues what AI answers say about a brand, with the original answers and sources behind every number rather than an opaque visibility score.

No. You do not need an AI provider key or subscription. Searches use your plan, top-up or earlier one-time pack credits, and IndexHalo pays for the free report you get after verifying your email. Every run shows its cost before it starts. Each answer keeps its AI source, question, timestamp and citations.

Yes. Draft a page in Optimize › Actions › Content to see its page checks: each reports pass, fail or unknown with the exact text it read, and nothing is weighted or added up. The checks read the page text; they do not predict whether an AI source will cite it, which only collected answers show.

Yes. Every plan can crawl same-domain pages across a site, using sitemaps and discoverable links, then produce a page inventory, site-wide metrics, failures, and prioritised actions.

Yes. Every plan can schedule daily, weekly or monthly recrawls and searches within the credit limits you approve, with dated snapshots. Change is compared only between matched repeat checks, and unavailable sources remain explicit.