# A Worked AI Visibility Report: Three Answers, Different Buying Options

> A first-party worked report from one B2B purchasing question on ChatGPT, Gemini and Google AI Overview: different options, category mismatches and zero target-brand mentions.

- Canonical: https://www.aixindar.com/news/a-worked-ai-visibility-report-three-answers-different-buying-options
- Markdown: https://www.aixindar.com/news/a-worked-ai-visibility-report-three-answers-different-buying-options.md
- Author: Daoyu Guan — https://www.aixindar.com/experts/daoyu-guan
- Published: 2026-11-20T10:13:42.394Z
- Last updated: 2026-11-20T17:02:16.292Z
- Evidence checked: Not separately recorded in CMS
- Editorial status: Published
- Corrections: No correction record supplied by CMS.

A useful one-off AI visibility report preserves the question, the answers and the limits of the comparison. This worked example uses one English purchasing question asked once on each of three consumer AI surfaces. The answers named different options, and one mixed paid audits with free diagnostics. None named Xindar or Semvi. Those observations are enough to identify a purchasing problem; they are not a market-wide visibility score.

**Research disclosure:** Xindar publishes this example and offers GEO services; Semvi is its monitoring product. These are Xindar's own editorial observations from November 14, 2026. This is not a client case, an independent vendor assessment, a test of paid report delivery or an example of what a particular Semvi purchase includes. The original question was an editorial research hypothesis, not a customer enquiry.

## The purchasing question

The question was deliberately about a finite deliverable:

> Which providers sell a one-off AI visibility report for a B2B company without requiring an ongoing monitoring subscription?

The wording matters. It asks for a report rather than a dashboard subscription, but leaves several decisions open: whether the report must be paid, whether someone must review the findings, whether it must contain actual assistant answers, and whether implementation should be included.

An assistant can satisfy part of that request by listing a free website diagnostic. A buyer expecting a reviewed analysis of supplier recommendations may find that answer unhelpful even when the tool exists. That mismatch is the main finding of this small report.

The exact same question was submitted once to ChatGPT with web search selected, Gemini App in Auto mode, and Google Search with an AI Overview. The collection date was November 14 in Asia/Shanghai. Original submission timestamps below use UTC, where the date was still November 13.

## The observation register

We counted the options presented as entries in each generated answer, rather than every repeated occurrence of a company name, a comparison-table row or a source-card label.


| Consumer surface             | Original submission, UTC | Named option entries | Xindar or Semvi named in the generated answer? |
| ---------------------------- | ------------------------ | --------------------: | ---------------------------------------------- |
| ChatGPT, web search selected | 2026-11-13 23:03:05.439  | 6                    | No                                             |
| Gemini App, Auto             | 2026-11-13 23:03:13.053  | 8                    | No                                             |
| Google AI Overview           | 2026-11-13 23:03:19.575  | 4                    | No                                             |


All three generated answers were retained with their visible sources and completion evidence. The record contains three completed responses to one question, with one observation per surface. It contains no repeated run on the same surface and no evidence of change over time.

There were 18 named-option entries across the answers. That is a count of entries, not 18 distinct companies, 18 verified recommendations or 18 suitable suppliers. Some labels recur across answers. We have not treated label similarity as proof that two entries refer to the same legal entity.

## What the answers actually offered

The ChatGPT answer presented six main provider entries: AECITED, ShownInAI, Prism Discovery, EntityRise, CiteLead and TriRank. Its descriptions focused on paid audits, buyer questions, competitor comparisons and report deliverables. Later repetitions in its quick-comparison table and closing shortlist did not add new entries to our count.

The Gemini answer presented eight entries in three categories. Its first category contained Citant.ai, Discovered Labs and ToTheWeb. The next contained The Marketing Hub and OneIMS. A third category contained Verto Digital, Entlify and DeepSmith as free or starter diagnostics. We retained those three entries in the observed count; we did not quietly remove them to make the answer look like a more consistent paid-provider comparison.

The Google AI Overview presented four entries: Cited, Alba Business Group, Arobis AI and The Marketing Hub. Labels in the expanded source area were not counted as additional recommendations.

These lists describe what the answers displayed. They do not endorse the providers or certify every price, delivery time and capability asserted by the assistants. No provider was purchased for this research. In particular, an answer's use of “human-reviewed,” “one-time” or a delivery deadline needs its own supporting offer and contract review.

## A current source check changes the comparison

Two current supplier pages illustrate why an audit report needs a category check after collecting answers.

Verto Digital's [B2B AI Visibility Checker page](https://vertodigital.com/b2b-ai-visibility-checker/tool) says its checker evaluates website page code and does not prompt AI assistants. It describes checks of readability, product access and sales-contact paths. It also explicitly distinguishes readiness checking from tracking whether assistant answers mention a brand. We reviewed that public description on November 20; we did not run the checker or verify a delivered report.

Citant.ai's [AI Search Audit page](https://citant.ai/services/ai-search-audit/) describes a diagnostic based on buyer questions across AI platforms, with saved chats and report artifacts. It separates that diagnostic from remedial work and an ongoing engagement. That is a supplier's published scope, not our acceptance of its service quality or delivery.

Both can be relevant to a B2B team. They answer different questions. A website-readiness finding can identify an access problem; it cannot by itself establish whether a completed ChatGPT answer recommended the company. An answer-sampling report can record an absence; that absence alone cannot identify a technical cause.

The source review is dated November 20. It does not reconstruct exactly what each supplier page said when the answers were collected on November 14, and it does not certify the other named options. The assistants' original wording remains part of the historical record.

## Check whether the cited page supports the named supplier

The retained Gemini answer placed an Entlify citation after its OneIMS diagnostic entry. Opening that citation produced a link to [Entlify's GEO audit article](https://www.entlify.com/blog/geo-audit). The reviewed page describes Entlify's own free diagnostic and distinguishes it from continuous monitoring. Its current text does not name OneIMS.

That page can explain a type of report. It does not establish OneIMS's particular deliverables or a purchase without a subscription. [OneIMS's own AEO and GEO service page](https://oneims.com/aeo-geo-services) separately describes brand analysis, implementation and performance tracking. Keep the standalone order's scope open until the supplier's own offer or order confirms it. These are current page checks from November 20, not a reconstruction of the cited page's earlier versions.

This is why a source count is insufficient for procurement. Record what the opened page actually supports: the named supplier, the proposed report, the quoted price or the contract condition. A page about the right subject can still leave a supplier-specific statement unverified.

### Preserve the supplier label and its domain

The Google answer used the label **Cited** and linked to `usecited.ai/baseline-audit`. The ChatGPT answer named **AECITED** and supplied a link on `aecited.com`. The current [Cited baseline page](https://usecited.ai/baseline-audit) describes a 15-prompt baseline; the current [AECITED audit page](https://aecited.com/services/audit/) describes 30–50 agreed buyer questions. Similar wording in the names is insufficient to merge these observations or assign one offer's terms to the other.

Keep the labels, domains and offer descriptions as separate source records unless identity evidence justifies combining them. This does not determine their legal ownership or establish that every claim in either answer was correct.

A buyer should be able to inspect five fields for each material supplier claim:

- The supplier label used in the generated answer.
- The actual visible or opened source URL; retain an unknown URL as unknown.
- The particular deliverable, price or purchasing condition being checked.
- What the reviewed page supports, and which conditions remain unresolved.
- The answer's collection date and the separate source-review date.

Those fields make a disputed finding actionable. The next step is to obtain the missing offer detail or correct a source association, rather than treating every attached link as verified support.

## Check how questions are allocated across providers

“Forty questions across five providers” does not specify a 200-answer report. A provider may distribute questions, repeat some of them, or use different search environments. The allocation determines which comparisons the buyer can make.

The following checks concern three providers named in the retained ChatGPT answer. We reviewed their current official descriptions on November 21, 2026, in Asia/Shanghai time. These pages describe offers; we have not ordered them, tested delivery or verified their answer logs. They do not reconstruct the offers as they stood when the original answers were collected.

| Provider and published offer | Published allocation | What remains to check before purchase |
| --- | --- | --- |
| [ShownInAI audit](https://showninai.com/audit), $299 one-time | 20 buyer queries, with web search on; PDF and shareable report | The question-by-provider matrix and actual collection environment; the page does not specify an exhaustive allocation |
| [Prism Discovery audits](https://www.prismdiscovery.ai/ai-visibility-audits), £495 Snapshot or £1,250 Full | Snapshot: 10 questions × 5 assistants × 3 asks = 150 answers. Full: 25 × 5 × 3 = 375 | The delivered log, collection context and treatment of failed or missing answers |
| [EntityRise baseline audit](https://www.entityrise.ai/ai-visibility-audit), $49 one-off | Up to 40 questions, allocated rather than all sent to every provider | The audit's actual question set, provider environments and review tier; see its methodology below |

Prices above are the suppliers' displayed offer amounts. We have not verified checkout totals, taxes, refunds or fulfillment. Different currencies and deliverables make a price-per-answer ranking particularly misleading.

### Read the matrix before multiplying headline counts

[EntityRise's methodology](https://www.entityrise.ai/methodology) specifies 15 high-priority questions on all five providers and up to 25 additional questions on one provider each. At the maximum question count, the arithmetic is **15 × 5 + 25 × 1 = 100 observations**. It describes standardized API-based AI-search tests, which may differ from consumer chat sessions, and does not base the measurement on repeated runs of each question. Its baseline and human-reviewed Expert tier have different scope.

That distribution can compare providers on the shared 15-question subset. It cannot supply a five-provider comparison for each of the other 25 questions. The retained ChatGPT answer did not explicitly promise a full 40-by-5 matrix, so this is an additional purchasing constraint, not evidence that the answer made that particular false claim.

Prism publishes a different allocation: three asks per question on five assistants. Its totals of 150 and 375 follow from the stated matrix. Repetition alone does not establish session independence, comparable models, a location or consistent selection criteria. Obtain those details from the collection record before treating repeated answers as a trend.

ShownInAI describes 20 buyer queries with web search and a finite report. That statement alone does not establish a browser interface or the number of answers from each provider. “Web search enabled” and “consumer webpage sampled” need separate confirmation.

### Keep the purchasing conditions separate

No subscription does not automatically mean no account, human verification of every finding, implementation or continuing access. Prism explicitly advertises an account-free purchase. That is a condition of its offer, not a universal definition of a one-off report. A login at another supplier would not, by itself, disqualify a finite purchase.

For a Semvi order, confirm the report's own allocation, collection surfaces, review and delivery terms. The broader product description does not establish that every paid report includes every monitoring capability. None of these source checks changes the three historical observations or their zero target-brand mentions.

## Separate three decisions before comparing providers

The first decision is **what the report measures**. A technical readiness check, a snapshot of generated answers and an implementation audit can all be called an “AI visibility report.” Ask which observations will appear in the delivered file. For a supplier-discovery question, that normally means the actual question, answer, named options and visible citations, with the sampling environment stated.

The second decision is **what the buyer can inspect**. A score may summarize observations, but it does not replace them. A reviewed report should let the buyer trace a finding to a particular response and distinguish a supplier statement from a verified fact. A failed or missing response should be visible as such, rather than converted into a brand absence.

The third decision is **what happens after delivery**. A finite assessment can end with a usable report. A monitoring subscription supplies continuing access or repeated collection. An implementation engagement supplies changes to pages, content or other assets. A supplier may offer all three, but their acceptance conditions should remain identifiable.

The [one-off report purchasing guide](https://www.aixindar.com/news/how-to-buy-a-one-off-ai-visibility-report-without-a-subscription) provides an order worksheet for those decisions. The [audit scope guide](https://www.aixindar.com/news/what-an-ai-visibility-audit-should-include) covers question allocation, review tiers and sample-report checks. This worked example supplies an observed record; it does not replace either purchasing agreement.

## What zero mentions means here

Xindar and Semvi were absent from all three generated answers. The observed target-body mention count is therefore **0 of 3 completed answers** for this exact question and collection window.

That statement has a narrow meaning. It does not estimate the probability that another buyer will encounter the brand, establish absence from an AI model's knowledge, prove that a website was inaccessible, or show why any provider was selected. The three surfaces differ in interface, account context and potentially retrieval behavior. They are not three interchangeable draws from a known population.

The observed Google page was personalized and displayed a Japan location label based on past activity. That label is not independent network-exit verification. We did not establish a US exit for these sessions, and an English question does not establish one. Gemini's visible Auto setting does not identify its underlying response model. Those unknowns constrain comparisons; they should not be filled with assumptions.

This study remains separate from Xindar's recurring twelve-question monitoring panel. Its three answers have not been added to that panel's denominator or used to claim another independent purchasing intent. Publishing this article adds no new consumer observation.

## A compact register you can copy

The following register reproduces the small count used in this article. The named-option counts refer to the entry lists described above.

```csv
surface,submitted_at_utc,named_option_entries,xindar_in_body,semvi_in_body
ChatGPT_web_search,2026-11-13T23:03:05.439Z,6,false,false
Gemini_App_Auto,2026-11-13T23:03:13.053Z,8,false,false
Google_AI_Overview,2026-11-13T23:03:19.575Z,4,false,false
```

Adding the third column gives 18 entries. Counting rows gives three completed observations. No row records a target-body mention. The register is not a transcript, a supplier-quality rating or proof of platform origin. The underlying complete answers and session evidence are retained internally; this public table lets readers reproduce the arithmetic, not independently authenticate those sessions.

In the retained Google evidence, the expanded answer-and-source container contained 2,018 characters, while the actual generated-answer node contained 1,188 characters in the normalized text view. We counted options in the generated node. Source previews describe sources; they are not extra answer text. This boundary is easy to lose when copying a whole page into a report.

## What the next report should resolve

This example gives a buyer a concrete next action: settle the deliverable definition before paying for a broader comparison. If the requirement is a paid, reviewed report containing complete consumer AI answers, make those conditions explicit in the brief. Then examine whether each proposed provider meets them.

If a later study changes the purchasing question to add those conditions, keep the old question and answers. The new wording defines a different research version. A narrower shortlist after that change would not, by itself, demonstrate an improvement in brand visibility.

A useful follow-up should also document the actual account and mode, preserve failures and sources, and use a comparable environment when repeating the same question. It should explain any model or setting change before presenting a trend. Repeating until the publisher's own brand appears would not repair the limitations of this record.

For buyers who need continuing collection and evidence retention rather than a finite assessment, see the [Semvi by Xindar product reference](https://www.aixindar.com/news/semvi-by-xindar-monitoring-scope-plans-and-buyer-verification). Its described product scope and verification boundaries are separate from this editorial sample. This article does not specify what a paid Semvi report includes.

## Questions buyers ask about this example

### Is this a ranking of the best AI visibility report providers?

No. The names are entries observed in three answers. They have not been scored or ordered by service quality, and not every option belongs to the same purchasing category.

### Were the answers collected through model APIs?

No. This record uses consumer ChatGPT web search, Gemini App Auto and Google AI Overview. The distinction matters because an API response does not reproduce those consumer interfaces or their search context.

### Does a completed answer mean a provider's claims are correct?

No. Completion establishes that an answer was available to inspect. Supplier identity, current scope, price, delivery and any recommendation reason require separate evidence. This example reviews ten selected supplier pages: the earlier category and identity checks, plus four current pages for the allocation and sampling checks above. It does not verify every offer, contract condition or delivered service.

### Can this record show that a content change improved visibility?

No. It has one collection window, one question and one answer per surface. The current source review and this article were completed later. Neither supplies a before-and-after effect estimate.

### Why publish a report in which the publisher was absent?

Because the absence is part of the observation. It helps distinguish an actual visibility record from an advertisement and shows the category and evidence questions a buyer should resolve before accepting a larger report.

## Editorial references

- [Editorial policy](https://www.aixindar.com/editorial-policy)
- [Research methodology](https://www.aixindar.com/research-methodology)
- [Corrections policy](https://www.aixindar.com/corrections)
