# 10 AI-Visibility Vendor Facts That Changed Since We Last Checked

> Source: CitedIndex — https://citedindex.com/blog/vendor-claims-that-changed-since-we-last-checked (sourced research, built to be cited)
> By CitedIndex Editorial · Published 2026-07-21 · Updated 2026-07-24

_Re-verifying listings before publishing caught 10 real pricing and feature drifts this month — what changed, why, and how to check a vendor's claims yourself._

Every listing in this index carries a `last_verified` date, and we re-check the number behind it before republishing anything that cites it — not on a fixed calendar, but whenever a listing comes up for a refresh or a page is about to restate one of its facts. Over the past few weeks that habit caught ten real drifts across nine different AI-visibility and GEO tools: a usage metric that had quietly dropped, a feature that moved to a cheaper tier, a pricing model that went from fully custom to partly self-serve, an engine-count headline that quietly disagreed with the vendor's own FAQ, an undercounted tracked-engine list, a tier-gating detail we ourselves got backwards on a first pass, and more. None of the ten involve a vendor being dishonest. All ten are proof that a SaaS product's pricing page, feature-gating and self-reported numbers — and our own first read of them — move faster than any secondary source, including our own past self, reliably tracks them.

That's the point of this piece: not to call out any one vendor, but to show buyers exactly how much a "current" fact can drift in a few weeks, with ten concrete, dated examples, and what to do about it.

## What changed

| Tool | What we had on file | What's true now | Kind of drift |
|---|---|---|---|
| [Knowatoa](/knowatoa) | A 7-day free trial (sourced from a third-party review) | No time-boxed trial of the paid product — a free, one-time, no-card AI-visibility audit instead | Secondary-source error, not a vendor claim |
| [GeoRankers](/georankers) | Grow tier: "2,100+ AI answers analyzed" | Grow tier ($129/mo): "1,100+ AI answers analyzed" | Self-reported usage metric |
| [Frase](/frase) | AI Visibility tracking gated to the $103/mo Professional tier | Included starting on the $39/mo Starter tier | Feature-tier gating |
| [Semrush AI Visibility Toolkit](/semrush-ai-visibility-toolkit) | Engine coverage stated without a tier boundary | Confirmed base tier ($99/mo): ChatGPT, Google AI, Gemini, Perplexity — Claude and Grok require Enterprise | Engine-coverage precision |
| [Profound](/profound) | No published pricing, fully sales-led | Self-serve Starter ($99/mo) and Growth ($399/mo) tiers now published; Enterprise stays custom | Pricing-model change |
| [Otterly AI](/otterly-ai) | API access unavailable | API and MCP access included from the $189/mo Standard tier up | Feature-availability claim |
| [Peec AI](/peec-ai) | Enterprise pricing and API access fully unpublished | A custom Enterprise tier (with API access) now sits alongside three published self-serve EUR tiers | Pricing-structure transparency |
| [Profound](/profound) | Enterprise tier: "up to 10 answer engines" | Enterprise tier: up to 9 answer engines — matching the 9 Profound names by name (ChatGPT, Perplexity, Claude, Gemini, Grok, Microsoft Copilot, Meta AI, DeepSeek, Google AI Overviews) in its own FAQ | Our own data error, not a vendor claim |
| [Promptwatch](/promptwatch) | Tracked ChatGPT, Gemini, Claude, Perplexity, Copilot and Google AI Overviews — 6 engines; no MCP server listed | Tracks 8 engines (adds Grok and DeepSeek) and ships an MCP server so Claude, Cursor and other MCP-compatible AI assistants can query its visibility, citation and mention data directly | Undercounted engine list + a missing feature |
| [Trakkr](/trakkr) | REST API gated to Scale as a paid add-on; MCP access gated to Scale (from our own earlier same-day check) | REST API bundled into Scale at no extra cost; MCP access ships from the Growth tier up — only white-label agency branding is a separate Scale add-on | Our own correction, made hours earlier the same day, was itself backwards |

Three failure modes are mixed together here, worth separating. Knowatoa and Profound are two of them: both errors were ours, made once. Knowatoa's was inherited from a third-party review that described a trial Knowatoa's own site has never advertised. Profound's was an internal inconsistency that predates this correction — our listing headlined "10" answer engines on Enterprise while our own FAQ entry for the same listing already named exactly 9 of them, and the live pricing page confirms 9 is correct. Neither is a vendor moving the goalposts; both are a reminder that a secondary source (including our own past self) can be wrong even when nothing about the underlying product changed. Trakkr is a third, more uncomfortable case: we checked trakkr.ai/pricing, corrected the listing, and got the correction itself backwards — REST API and MCP tier-gating reversed — discovered only when a second, unrelated task (drafting a new comparison page) sent us back to re-read the same live page a few hours later and four independently-worded fetches all disagreed with our first pass. A single live check is not automatically ground truth, even a same-day one; it's one read of a page that can itself be misread. The other seven are the more common case: the vendor's own product genuinely changed. A tier got a new feature, a usage cap moved, a pricing page that used to say "contact sales" now has real numbers on it, or — Promptwatch's case — the product quietly added engine coverage and a whole new integration surface that our copy hadn't caught up to yet.

## Why vendor facts drift this fast

None of the seven product changes above came with anything resembling a changelog. That's normal for this category, not a red flag on any one vendor:

- **Tiers get renegotiated as a product matures.** A feature that shipped as a Professional-only differentiator six months ago (Frase's AI Visibility tracking) can migrate down to the entry tier once the vendor decides it's now table stakes, not a premium hook.
- **Self-reported usage numbers move with infrastructure cost, not marketing intent.** GeoRankers' "AI answers analyzed" cap on its Grow tier is exactly the kind of number that gets adjusted as LLM API pricing and rate limits shift underneath the product — up or down, with no announcement either way.
- **"Contact sales" is a starting position, not a permanent one.** Profound and Peec AI both moved from fully custom/opaque pricing toward a published self-serve floor while keeping a custom top tier — the more common direction for a vendor chasing self-serve volume once it has enough market validation to publish a number.
- **Feature gating is invisible until someone checks the actual tier, not just whether the feature exists somewhere.** Otterly AI's API access was never absent from the product — it was gated to a tier a page-level glance ("does this tool have an API?") answers wrong just as easily as right.
- **"Which engines does it track" is a moving target, not a fixed spec.** Promptwatch quietly added Grok and DeepSeek tracking, plus an entirely new MCP-server integration, without any of it showing up in a headline announcement — the only way to catch it was checking the product's own docs page directly, not the marketing homepage.

## What this means if you're the one buying

A directory, a review site, or a comparison article is a snapshot, not a live feed — including this one. Three habits protect you regardless of which source you're reading:

1. **Check the live pricing page yourself before a purchase decision, not the page that convinced you to look.** A review or a directory entry that's a few months old is exactly old enough for one of the drifts above to have happened.
2. **Ask which tier a feature lives on, not just whether the tool "has" it.** "Does it have an API" and "does the tier I can afford have an API" are different questions, and vendor marketing copy tends to answer the first one loudly and the second one in the pricing table's fine print.
3. **Treat a vendor's own usage/volume numbers ("N answers analyzed," "N prompts tracked") as a snapshot of today, not a spec.** They're the number most likely to move without any public notice, because they're operational, not structural.

## FAQ

**Does this mean the vendors above were misrepresenting their products?**
No. Seven of the ten changes are the product genuinely evolving — a tier gaining a feature, a pricing page publishing a number that used to be custom-only, an engine list quietly growing. That's normal SaaS iteration, not misrepresentation. The other three (Knowatoa's trial claim, Profound's engine count, Trakkr's API/MCP tier gating) were our own errors — one inherited from an outdated third-party review, one an internal inconsistency between our own headline figure and our own FAQ, and one a same-day correction that got the tier gating backwards on its first pass — not anything any of the three vendors themselves claimed.

**How often do you re-check a listing's facts?**
Not on a fixed schedule — whenever a listing is due for a refresh, or before any other page on the index restates one of its numbers. Every correction above carries a dated source and updates the listing's own `last_verified` field, which you can see on each tool's page.

**What do you do when you can't confirm a claim?**
We leave it out, or say plainly that we couldn't verify it, rather than publish a number we can't currently trace to a live, primary source. That standard is documented on our [methodology page](/methodology).

**Is any AI-visibility directory's data ever fully current?**
No single source is — that's the actual argument of this piece. The realistic goal isn't a permanently correct snapshot; it's a short gap between when a fact changes and when the next check catches it, plus giving the reader the tools to verify the number that actually matters to their decision.

## Methodology & sources

Every figure above is drawn directly from the listed tool's own live pricing or product page, re-verified between 2026-07-21 and 2026-07-24 and reflected in that tool's listing on this index — see each linked listing for the full source citation and verification date. This is the same live-source standard applied to every fact in this index; see our [methodology](/methodology) for how listings are researched and gated before publishing.

Primary sources checked directly: [Knowatoa — pricing & onboarding](https://knowatoa.com/onboarding/start) · [GeoRankers — pricing](https://georankers.ai/pricing) · [Frase — pricing](https://www.frase.io/pricing/) · [Semrush — AI visibility pricing](https://www.semrush.com/pricing/ai/) · [Profound — pricing](https://www.tryprofound.com/pricing) · [Otterly AI — plan details](https://help.otterly.ai/pricing-of-otterlyai) · [Peec AI — pricing](https://peec.ai/pricing), Enterprise-tier detail via [OMR Reviews](https://omr.com/en/reviews/product/peec-ai/pricing) · [Promptwatch — prompt tracking](https://promptwatch.com/prompt-tracking) and [MCP server docs](https://promptwatch.com/docs/mcp/introduction) · [Trakkr — pricing](https://trakkr.ai/pricing), re-checked four separate times the same day before the tier-gating detail was confirmed correct.

To see current, dated pricing and feature facts for any tool in this index, start with [the index](/) or [best AI-visibility tools](/best-ai-visibility-tools).

---

Tags: geo, ai-visibility, methodology, pricing, editorial