What changed,
in public.
Every material product and measurement change, dated and newest first. Entries keep their original dates; when something was later replaced, the entry says so instead of disappearing. For what IndexHalo does today, see Capabilities.
Product changelog
| Date | Change | What it means |
|---|---|---|
| You against your rival, no account | Type your website on the homepage, and the rival you lose deals to, and see in seconds which of you each of the six AI sources can read. Then ask all six who they name: your free report starts as soon as you're signed in. | |
| Every brand's own mark | The workspace reads each brand's icon from its own website, once, on the server, and shows it in the brand switcher and search. A brand whose site has no readable icon keeps its initial; nothing is guessed from a name. | |
| Named against recommended, by AI source | Insights › AI sources and the Overview's What each AI said show how many of each source's answers recommend the brand, outright or for a stated case, beside how many name it. The Executive report gains a What each AI source does table and lists the sites the answers cite most with whose each is. | |
| Clearer rankings, cards and gap analysis | A product category written like a name ("sales CRM") is no longer ranked as a brand. Ranking shows each brand's net tone instead of a badge that read the same for all. The Overview's What changed card states the change. Gap analysis lists the AI sources that left you out as one card per question. Reports has one Run again control. | |
| What the AIs say, named against recommended | The homepage shows, from the published example, how often each AI source names a brand and how often it recommends it, who is named in its place, which buyers each AI sends where, which cited sites go with being recommended, and where two AIs disagree. | |
| Authenticator-app sign-in | Any account can add a code from an authenticator app to its password sign-in, with ten one-time recovery codes for a lost phone, in Account and data. An organisation owner can require it for every member in Settings › Team. Google and Apple sign-in keep their own checks. | |
| A month's report, published | Pricing, the sample report and the comparison guides link a month's report built from the read-only example's 63 answers to 8 questions in 2 markets, with what changed between checks. | |
| Coverage by question and AI source | Coverage lists each question once, with a column for each AI source, and one row per market and device. Times on the page are in UTC, as in the filters. | |
| Every answer counted in Chats | The figures above Chats count every answer in the selection, the same answers the answer log and Export count, not only the first page loaded. | |
| Your plan in Settings | Settings › Billing names your plan, its monthly credits and its included weekly re-checks, with the other plans beside it. | |
| Compared brands in search | ⌘K search finds the brands you compare against, as well as pages and saved reports. | |
| Full answers as PDF and Word again | Every full answer in a report downloads as a PDF or Word file again, quickly. | |
| Reports read question by question | Executive reports give each question its own section: what each AI source answered, the pages it cited, linked, and what to do for that question. | |
| Answer replay | See one question's answers over time, and what changed between them. | |
| Repeats and model changes | All prompts shows how often each question was asked. Each answer names the model its AI source reported, and charts mark when that model changed. | |
| Why that page was cited and yours wasn't | Put a page the answers cited beside your own and compare them side by side. | |
| Merge duplicate brands | My websites merges two copies of the same brand into one, keeping all their evidence. | |
| A faster, steadier workspace | The workspace loads sooner and no longer shifts while it loads, and the longest analyses run in the background. | |
| Who replaces you | Gap analysis › Who replaces you shows which brand takes your place when an answer leaves you out. | |
| Positioning map | Ranking › Positioning map shows which brand the answers recommend for which buyer. Actions and reports carry it too. | |
| Trash, Archives and Delete permanently | Trash and Archives sit in the sidebar. Brands, saved prompts and items in Trash can be deleted permanently, by you. | |
| Source influence | Cited sources › Source influence shows which cited sites go with your brand being recommended. | |
| Search to answer | SEO › Search to answer puts Google's top 10 results for the searches an AI ran beside the pages its answer cited. | |
| Question generator | About 20 buyer questions from one keyword, in the chosen market's language, from a monthly allowance that uses no credits. | |
| Whole audits in one step | Paste up to 200 questions at once, see the duplicates before saving, and agree one quote for the set. | |
| Recheck after publishing | When you mark a page published, IndexHalo offers a quoted check of the same questions 14 days later, or your plan's included weekly re-checks. | |
| Searches recover by themselves | A search that stalls is recovered automatically; nobody has to start it again, and a search that cannot finish uses no credit. | |
| A fuller example | The read-only example holds eleven real HubSpot searches, and the 20 A–Z dashboard lessons sit on /academy/ and beside each feature on /features/. | |
| Recheck now and a reminder after publishing | Recheck now on a task or a saved question set shows the quote first, then asks the same questions again with the same sources, countries and languages, so only matched answers are compared. After you publish a page change, Remind me to recheck sets a reminder 14 days out. | |
| Included weekly monitoring | Professional includes 10 questions re-checked weekly; Business 30; Agency 150. No credits are used. You choose the questions and agree to the exact terms first, in Reports › Schedules; monitoring can be paused and resumed. | |
| Classified by you | Answers that name your brand without a clear recommendation can be classified by you: recommends it, recommends it for a stated case, names it without recommending, or advises against it. Each verdict is dated with who made it, counted apart from IndexHalo's own figures and never added to them, and can be withdrawn. A site credited to a competitor only by its name reads Unconfirmed until you confirm it or mark it as not that company. | |
| Is this the company you mean? | The Overview flags answers that cite a site named like a compared brand but not its website. Confirming a different company is dated and can be undone, and the affected questions can be re-run once free, with a dated correction notice on the original report. The evidence explorer gains Something looks wrong here, and reports gain an optional Who we compared section. | |
| Question types, issue register and digest | Report tables gain a Question type column. The reputation issue register downloads as CSV or Excel, and the digest reports presence by question type. | |
| Free product check and branch import | Check my products for free matches a product file against the answers you already have, without saving the catalogue, running a search or using a credit. Branches can be imported from your own website, and Actions show the effort each task takes. | |
| Agency controls | Monthly credit budgets per client or brand group (credits, never money), optional client sign-off of the comparison set before the first paid run, agreed next report dates, and 30-day pitch workspaces kept out of client counts until promoted. | |
| Teams and security pages | New pages for in-house teams and for security, with a security.txt contact. Both say plainly what is not available yet: single sign-on (SSO), SCIM provisioning and multi-factor sign-in. | |
| Website facts from the product's own code | New capabilities and integrations pages, rewritten API and MCP documentation, a dated changelog and corrected FAQ, pricing and article copy. Every count and list on them is read from the code, so the site cannot drift from the product unnoticed. | |
| Each question keeps its own market and language | Every saved question keeps its own country, or a region or city for the Google sources, and its own language, and Copy to markets adds it to other countries without translating its wording. Runs, schedules and repeat checks use each question's own market and never fall back to a default. Ranking, the Overview briefing, Present mode and the PDF, Word and PowerPoint reports split results by market, each answer names its market and language, and dates and numbers appear in one English format. Measurement change: fact review, page types, the tone of linked passages, the Sentiment Shield and narrative incident review now read German, French, Spanish and Dutch answers too, so figures that include those languages can differ from before this date. | |
| Competitor websites, language and region from your input | Competitors are matched by their own website, with a free “Find their website for me” preview, and competitors saved by name only are flagged in reports. The search language follows the language of your questions, and the market follows your choice and your website's own signals; setup says why. | |
| Free re-run when a search measured a namesake | When answers described a different company with the same name, confirm it in the Overview and save the competitor's own website: the affected questions in that search can be re-run once at no charge, after you see the quote and agree. The credits come back as their own labelled credit entry, never as a payment or refund. | |
| Answer tone in the data archive | Answer records exports (CSV, TSV, XLSX, JSON and NDJSON) gain a tone column with the answer log's label for your brand in each answer: positive, neutral, negative, mixed or unmeasured, blank where your brand is not named. It replaces the numeric sentiment score, which is also gone from the legacy Looker feed. | |
| German, French, Spanish and Dutch answers classified | Tone, recommendation, reputation and positioning findings now cover German, French, Spanish and Dutch answers as well as English. Answers in other languages show “not classified (language)”, never a zero. Measurement change: figures that include these languages were unclassified before this date. | |
| Brand shape | One two-ended descriptor model (for example affordable or premium) replaces the radar chart in the dashboard and in reports, with the answer counts in every cell. | |
| Role homes and Why this answer | Overview's “View for” opens homes for Content, SEO & web, Brand & PR and Product marketing. Why this answer shows the searches an AI ran, the pages it read and the passages it cited. | |
| Four more markets and no composite scores | Austria, Switzerland, Belgium and New Zealand join the markets. Composite scores out of 100 are replaced by per-check findings with their denominators. | |
| BI feeds, Looker Studio and scoped MCP | CSV, JSON and NDJSON feeds for Google Sheets, Looker Studio, Excel and Power BI; a Looker Studio Community Connector as Apps Script source, shared on request; an account-wide export; polling endpoints for Zapier, Make and n8n; and an MCP endpoint that accepts scoped API keys only. | |
| CMS, CRM and project tools | Drafts and summaries to WordPress, Webflow, Notion and HubSpot, and evidence-backed actions to Asana, ClickUp, monday.com, Jira and Trello. Publishing stays a step a person approves. | |
| Upgrades, top-ups and brand GA4 import | Self-service upgrades with a prorated preview, top-up packs at plan prices that expire with the plan credits, a GA4 import per brand for AI referrals, and Repeat this check for the same questions under the same scope. | |
| Monthly plans | Professional $49 for 100 credits, Business $129 for 300 credits and Agency $299 for 750 credits a month, with the same features on every plan. One credit is one complete search across the selected AI sources. | |
| Integrations platform | Scoped, revocable API keys, REST API v1 with an OpenAPI description, signed webhooks, Slack and Microsoft Teams messages, and an in-app inbox. | |
| Organisations and roles | Members with roles, brand-limited access, approvals for runs above a credit threshold, and an activity log. | |
| GA4 and Search Console sync (built) | Read-only GA4 and Search Console sync beside measured visibility. It is built and available once connected. | |
| Scheduled report delivery | Saved reports emailed on a daily, weekly or monthly schedule to the people you choose, with a dated notice when the data is not fresh. | |
| One-time credit packs (replaced) | One-time credit packs and an auditable credit ledger. Replaced by monthly plans on 18 October 2026. | |
| Design review implemented end to end | The landing page split into focused URLs (/method.html, /sample-report.html), the signed-in Intelligence Hub gained real routes (/app, /app/reports/:id, /app/sites/:id, /app/studio), the dashboard went delta-first with an activation checklist and actionable failure rows, and reports now lead with the fix, keep the score on screen while scrolling, and print self-explanatory. | |
| Shared design tokens and view primitives | One token file now feeds the marketing pages, the Intelligence Hub, and the editorial pages, so a colour or type change cannot drift between copies. | |
| 87 measured AI crawler reports published | Fifteen AI user-agents resolved against the live robots.txt of 87 well-known websites — every verdict dated and traceable to the deciding rule. Public at /ai-crawler-report/. | |
| Sourced competitor comparisons | Three head-to-head pages that state where the competitor is the better choice and link every claim to the vendor's own pricing page. | |
| Citation Studio | Block-based composer for citation-ready pages: fourteen measurement-backed block types, live scoring as you type, and export to semantic HTML, JSON-LD, llms.txt and robots stance. | |
| Answer-engine intelligence panel | Persona stress tests, defensive probes, citation poaching and parentage, cross-language drift, fair-use audit, and the model-version regression radar across four measurement phases. |
Entries before 18 October 2026 describe earlier versions of IndexHalo. Measurement-affecting changes, such as a newly classified language or a changed definition, are called out in the entry itself, because a figure that moved because the ruler changed says nothing about your content.
AI sources
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. Each retained answer records the source, market, language, collection time and the model version where the source reports one.
| Source | How it is collected | In every search | What you get |
|---|---|---|---|
| Google Organic | Search results page | Separate search, one credit | Organic search results. Kept separate from AI-generated answers. |
| Google AI Mode | AI search answer | Yes, by default | A captured AI Mode answer and its available sources. |
| Google AI Overviews | AI search answer | Yes, by default | The AI Overview Google showed above its results for the question, with the sources it cites. It usually arrives after the other answers and is added to the report when it does. If Google shows no AI Overview, the report says so. |
| ChatGPT | AI assistant answer with web search | Yes, by default | A captured ChatGPT answer with web search requested. Citations are not guaranteed. |
| Gemini | AI assistant answer with web search | Yes, by default | A captured Gemini answer and its available sources. |
| Claude | Model API answer with web search | Yes, by default | A Claude model API answer with web search requested, retained with its citations and collected by a background job. This is distinct from a signed-in Claude conversation. |
| Perplexity | Model API answer with web search | Yes, by default | A Perplexity Sonar API answer and its citations, collected by a background job. This is distinct from a signed-in Perplexity conversation. |
Proof you can check
- Measured AI crawler study — 30 AI user-agents resolved against the live robots.txt of 389 well-known websites, dated, with the deciding rule behind every verdict.
- The real business report — the complete retained HubSpot search, including source answers, citations, competitors and the next actions.
- The method — how answers are collected and compared, and what IndexHalo measures directly, what it cross-references, and what it only estimates.
SEE THE INTERFACE
Open the complete business report →Judge the structure, not the logo wall.
The real HubSpot search shows how a complete report connects observed answers, citations, competitor comparisons and next actions.