Changelog

The pricing page now shows a side-by-side table of what each plan includes — brands, prompts, competitors, scan frequency, samples per prompt, engines, access, reports and exports — and explains why we track 30–40 prompts. A new changelog page (with an RSS feed) lists dated product and methodology changes.

The methodology page explains why we measure through official APIs rather than by scraping chat interfaces — and what that means: API answers can differ from what a signed-in user sees in the app. It also describes how tone and warnings are measured and where those labels stop being reliable.

One-off audit reports now include an «AI readiness» check of your website: whether AI search and assistant bots may read it, whether it has an llms.txt file, which structured data it publishes, and whether canonical and language links are in place — each with a short recommendation.

Competitor comparison pages were re-checked against each vendor's own pages: contradictory or outdated statements were corrected, claims that could not be verified were removed, and every page now shows its sources and the date they were verified. The privacy policy now describes exactly what happens to the email address you send through our forms.

Engine requests now carry your project's country. ChatGPT search, Perplexity and Google AI Overviews answer as they would for a user in that country — before, some answers assumed the user was in the US. Gemini offers no location setting for search grounding, so its answers are unchanged. For projects outside the US, compare trends before and after this change with care.

Recommendations are now created automatically when a scan finishes — no report needed. Items you marked as done or dismissed no longer come back reworded, and when a recommended page already exists on your site the suggestion becomes «Strengthen» that page instead of «Create». Alerts now name the engine that is failing, and an engine that stops answering raises a critical alert even when the others are healthy.

The competitors page separates possible competitors from the publications that are cited most, so news sites and directories are no longer suggested as competitors. Competitor domains are cleaned up automatically when added (no more https:// or trailing slashes). The prompts CSV export can be imported back without losing or mixing up columns, and cells are protected against spreadsheet formulas. PDF downloads get a readable file name with the client name and date.

In reports, answers given from the engine's memory are no longer labelled as errors — they are shown as their own class. When an audit has no branded questions, Brand Knowledge now says so instead of showing a bare dash.

The newsletter sign-up is now recorded separately from sales enquiries. The mobile dashboard menu and the home-page contact form work better with keyboards and screen readers.

An answer that our analyzer could not process is no longer counted as «your brand did not appear». It is now shown as not measured and left out of every figure — dashboard, report and recommendations use exactly the same answers. When it happens, the report says so openly: «N answers were not analyzed — not counted».

Google AI Overview answers are now read in full, including nested lists, expandable sections and tables. Before, a brand named only inside a nested list could be missed. AI Overview figures in new audits may shift slightly; the formulas are unchanged and earlier scans keep their original values.

Product cards shown inside a Google AI Overview now count as part of the answer, labelled as a separate surface. Only the product names are read — prices and ratings are not. An AI Overview that consists only of product cards is now counted as an answer, so AI Overview figures in new audits may rise slightly for shopping-style questions.

A finished scan no longer means «report delivered». Each audit now shows a delivery checklist — scan, analysis coverage, metrics, report, PDF, quality check, delivered — and a failed PDF or metrics step can be repaired without re-running (and re-paying for) the scan.

Quality-check and delivered marks can now be undone with a written reason, and every mark, undo and coverage confirmation is kept in a history with who did it and when. When some answers could not be analyzed, a report can only be marked delivered after confirming that the coverage was reviewed. Older reports that were created without a PDF now show the PDF as unavailable on the Reports page instead of looking complete.

A scan stopped while its answers were being analyzed could still end up marked as completed. A stopped or failed scan now keeps its real status, and its partial results are not counted in your daily figures.

The sample report linked from the pricing page is now a newer one, measured after the 28 September change to Brand Knowledge. The previous sample stays online because a blog post analyses it.

When a scan finishes, a notice now shows which AI engines answered in full, which answered only in part and which did not answer, with the reason: no response, timed out or blocked. It also says how many of the scheduled questions got an answer.

Brand Knowledge now counts only answers that actually recognize your brand. The analyzer reads the question as well as the answer, so a reply that repeats your name but says it can't find the brand, or confuses it with another company, no longer counts as an appearance. Expect lower Brand Knowledge figures from this date — a step in the trend line marks the change in measurement, not a drop in your visibility. Earlier scans keep their original classification. Discovery Visibility can shift slightly; citation figures are unaffected.

The privacy policy now lists TypeSafe, which independently checks how AI answers are classified; it receives project prompts and AI answers, never account data.

English pages now live at the root of tamgio.com — for example tamgio.com/pricing instead of tamgio.com/en/pricing. Every old /en address permanently redirects to its new location, so existing links and bookmarks keep working; the Azerbaijani version stays at tamgio.com/az.

The closing call-to-action no longer reads «Millions of questions. Millions of users.» — a line that could be taken as a claim about TAMGIO's own usage. It now describes the market: people ask AI assistants millions of questions every day. Three guide and use-case answers that described what «TAMGIO users» or «most TAMGIO projects» do now describe a typical setup of 30–40 tracked prompts instead.

The one-off audit scope is now written in one place and matches what is actually run: 10–12 prompts, each asked twice on three AI assistants, plus Google AI Overviews, delivered as a PDF and an unlisted link. Setting up an audit now suggests 12 prompts instead of 20.

The contact form no longer reads like a sign-up for a future launch. It is a request for an audit or a question about monitoring, and it says plainly that prices are public and that we reply by email with the scope.

The pricing page now answers questions that AI assistants could not answer about TAMGIO in our tests: cancelling (any time, no long-term contract), when an audit is refunded, deleting your data on request, and how payment works without a self-serve checkout.

AI answers in the dashboard, the shared report and the PDF are now shown with their formatting — headings, lists, tables and bold text — instead of raw markdown symbols.

The dashboard redesign is complete: every page now shares one layout with clear stat tiles and section icons. Only the look changed — the data, formulas and actions are the same.

63 more English reference pages — glossary terms, prompt libraries, engine pages and alternatives — each linked to related pages.

Starting an audit from the top bar now opens a confirmation with the plan, the estimated cost and time, and a note that a new report will be created while the existing shared link keeps working. The project picker has search and separate groups for monitoring projects, one-off audits and paused projects.

Prompts now keep their history. Editing a prompt saves it as a new version and keeps the previous text, and a prompt that already has results can be archived but not deleted — the evidence behind past figures is never lost.

If the domain set for a project is never cited while a near-identical domain is, the dashboard and the report now say so and show how often the other domain was cited. The domain can be corrected in Settings, and citation flags are recalculated from the stored answers.

When creating a project fails, the message now says why. A new project is saved in one step — its prompts, competitors and brand names are stored together or not at all, so a failed attempt no longer leaves a half-created project.

English is now the main language of the site: tamgio.com opens in English, and a flag switch (EN · AZ) in the header, mobile menu, footer and dashboard changes the language.

A competitor now counts as mentioned only when its name or domain actually appears in the answer text or its citations. Earlier analyses were re-checked with the same rule, and the affected reports carry a note saying so.

A new project starts with an explicit choice between monitoring and a one-off audit. AI prompt suggestions are based on your home page rather than the brand name alone, and «Find competitors with AI» uses a live web search, leaves out generic sites and suggests only domains that respond.

Reports and PDFs now carry the TAMGIO logo, who prepared them and the date, and audits are marked as one-off audits. A shared report link can be revoked — the page and its PDF stop working immediately.

The illustrative dashboard on the home page now shows a sample visibility of 38% instead of an eye-catching 78%. It stays labelled as sample data, not a measurement.

A public contact address, hello@tamgio.com, is now shown on the About page, in the footer and in the privacy policy.

The example answer in the home-page hero was presented as a real AI answer. It is now labelled as an example, and the panel carries a visible «Illustrative — sample data» caption.

Prices have their own page at tamgio.com/pricing. Before, they were only a section of the home page, so AI assistants looking for a pricing page found nothing.

New English reference sections: a glossary of AI-visibility terms, prompt libraries by industry, a page for each AI engine — including the ones TAMGIO does not measure, which say so — and pages on alternatives to other tools.

New About and Methodology pages: who builds TAMGIO, and how every figure is measured — the steps, the answer statuses and the formulas.

Report headline figures could show a dash even though the answers had been measured, because daily totals were calculated a moment before the scan was marked complete. Totals are now calculated after completion, and the missing daily figures were rebuilt from the stored answers.

Buttons on the site no longer promise things TAMGIO does not offer — a no-cost analysis, a trial without a card, a demo video or instant sign-up. They now lead to what exists: ordering a one-off audit, with setup done by us.

A new design for the home page, blog and guides, with one shared header and footer on every page, a mobile menu and a cookie-settings button in the footer.

A cookie banner now lets every visitor accept or decline analytics cookies, and the choice can be changed later from the footer. Advertising cookies are never used, and analytics runs only on the public site.

Prices are now public: three monitoring plans and a one-off audit, setup included, shown on the home page and on every comparison and use-case page.

Comparison pages now show when each competitor fact was last checked and link to the competitor's own pages. The site also stopped calling TAMGIO a self-serve tool: there is no sign-up form, and setup is done for you.

New overview pages for guides, comparisons and use cases, and a privacy policy that describes which data the site and the product handle.

Daily figures now include only scans that finished completely. A stopped or failed scan covers only part of your questions, so it is left out instead of skewing the rates; the formulas did not change and no past figures were affected.

A new Projects page lets you deactivate, archive and restore projects. Archiving hides a project from lists without deleting its data, and a restored project comes back deactivated.

The scan-finished alert now appears within seconds of a scan ending — before, it could take up to an hour. A scan indicator on every dashboard page shows scans in progress, and each project card shows whether it is queued, running or just finished.

«Scan all» queues every eligible project with one confirmation, scans can be started from the project cards on the «All projects» page, and progress appears as soon as a scan starts.

A running scan can now be stopped from the dashboard, and the number of samples per prompt can be changed in project settings. If a scan has to retry, the progress shows which attempt it is on instead of silently starting again from zero.