First count the outcomes of collection
The AI Overview route retained 540 attempt records across 20 question groups. The saved outcomes were:
Read each row with its scope and evidence.
| Collection outcome | Records | What the count means |
|---|---|---|
| Usable saved answer | 364 | Nonempty extracted answer text, with no retained error flag under this study’s classification. “Usable” does not certify factual accuracy. |
| Unavailable capture | 175 | Failed attempts with retained error evidence. These are not answers saying the brand was absent. |
| Valid no-overview observation | 1 | Successful saved search evidence with no overview and no retained nested error. This remains separate from both failure and a usable answer. |
| Total attempts | 540 | The three outcomes above combined. An attempt is not necessarily one upstream provider request or one billed call. |
The collection outcomes reconcile to that total: 364 + 175 + 1 = 540. It describes the retained evidence from this route and run, not what every person using Google experienced.
Next, we compared the text in the 364 usable records. Within each exact question group, the first occurrence of a text counts once. Further records with that identical saved text count as additional occurrences. Summed across the 20 question groups, there were 205 distinct strings and 159 additional exact-text records.
That is a second calculation on a smaller population: 205 + 159 = 364. The 205 count is not a fourth collection outcome and should not be placed beside failures as though it were one.
The same 48 attempts left different evidence to read
Table row 1: Q7, targeted unbranded. Table row 2: Q3, explicit lookup control. Different tasks, not matched experimental variants.
The two examples below were selected to make the contrast visible. They are different questions, including different study cohorts, rather than a controlled experiment.
Scroll horizontally to compare every column. All counts and questions remain available below.
| Saved question | Attempts | Usable records | Unavailable | Distinct saved text strings |
|---|---|---|---|---|
| After an acquisition, how should a website explain the relationship between the parent company and a product that still uses its old name? | 48 | 47 | 1 | 13 |
| Find Gist’s full Halloween research report on gistgeo.com. What did it find about renting an elaborate costume for one night, and what were the collection date and answer counts for that comparison? | 48 | 22 | 26 | 22 |
For the acquisition question, one exact saved answer text appeared in 22 of the 47 usable records. Its first and last occurrences were recorded about nine minutes apart. The text and citation list in those records match. The archive establishes repetition; it does not identify whether caching, retrieval behavior, extraction or something else produced it.
For the explicit report lookup, all 22 usable texts differ exactly. That does not establish 22 different conclusions. Two complete examples discuss the same costume-rental comparison with different wording and detail; one contains an incomplete bullet. A nonempty, distinct string is not automatically a more accurate or complete answer.
The contrast also prevents a misleading interpretation of the first row. Repeated text was not universal: the report-lookup question and one other question had no exact repeats among their usable records. Across the full set, 18 of 20 question groups contained at least one repeated saved text.
Keep the original question groups visible
The diagnostic included five explicit Gist or site-lookup controls and 15 deliberately targeted unbranded questions. The prompts were selected for a retrieval diagnostic, not sampled to represent market demand. Combining them can describe archive volume, but it cannot turn the run into a general visibility benchmark.
Scroll horizontally to compare every column. All counts and questions remain available below.
| AIO question cohort | Attempts | Usable records | Unavailable | Valid no-overview | Distinct saved texts within questions | Additional exact-text records |
|---|---|---|---|---|---|---|
| Five explicit lookup controls | 196 | 92 | 104 | 0 | 62 | 30 |
| Fifteen targeted unbranded questions | 344 | 272 | 71 | 1 | 143 | 129 |
| Total | 540 | 364 | 175 | 1 | 205 | 159 |
The cohort differences do not establish that adding a brand name causes failures or prevents repetition. They preserve the design so readers can see what the totals combine. The accompanying data table retains every exact question, its cohort and its counts; the two examples do not stand in for the whole panel.
Our earlier report on lookup versus unbranded discovery examines owned-link presence and claim accuracy in this study. Those are separate analyses. A count of distinct text strings does not say which records cite Gist, mention a product or recommend a business.
What “distinct text” means here
We compared the archived extracted answer-text field exactly as saved, using its UTF-8 representation. We did not trim spaces, change capitalization, normalize punctuation, group similar meanings or combine different questions. A text repeated under one question contributes once to that question’s distinct-text count, even if it appears in many retained records.
This is not a comparison of full raw provider payloads, screenshots or every retrieved URL. Two equal text fields may have differences elsewhere in their records. Two unequal strings may differ only in formatting or a small wording change. The count measures exact textual recurrence, not independent retrievals, semantic diversity or effective statistical sample size.
The archive’s attempt identities are unique and the study was admitted once, with no manual repeat job. That distinguishes this recount from accidentally importing the same record twice. It does not establish independence among upstream responses.
The computation covers every usable text in these 20 AIO cells. Four complete selected records were read for the examples; this is not a fresh factual audit of all 364 answers. The data and selected full texts are available below so the counting rule and examples can be inspected separately.
Report the evidence you have, without silently replacing it
Keep repeated records if the report’s purpose is to describe what was retained across attempts. Present a distinct-text view separately when the question is how much different saved wording an analyst has to inspect. Deleting repeats without explanation would change the analysis population; treating every retained text as an independent observation would claim more than this archive establishes.
The useful result is three clearly named counts: attempted collection, usable saved answers and exact text recurrence. In this study they were 540, 364 and 205 distinct strings within question groups. A report that names those units gives the reader a way to assess its evidence. A single unlabeled “sample size” conceals the differences.
Evidence and methods
The public appendix contains the 20 exact questions and their route-specific counts, plus four complete selected saved texts. The full study covered 60 question–route cells, but the collection-outcome and text-recurrence results in this article are AIO-only. Collection ended October 29; this later recount did not make new AI requests. The source archive and failed attempts remain preserved.
The 175 unavailable attempts all retain error evidence. The one valid no-overview observation has successful saved search evidence and no nested error. These classifications were checked separately before counting distinct text. Exact string comparisons and saved hashes were verified against the archived records; no semantic similarity threshold was used.
Thirty-five of 383 generated commentary rows in the wider study had blank question text. This recount uses the preserved question identities and archived extracted answer-text fields, not those commentary summaries. It makes no claim about why the route repeated text, why a source was selected or whether a website change caused a result.
Download the 20-question data as JSON · Download CSV · Inspect four complete selected saved answers. The selected answers are unedited generated outputs for evidence inspection; their advice and factual claims are not endorsed.
Inspect the complete public evidence
Read all 20 exact questions, cohorts and counts
Question groups are not matched experimental variants. Counts below cover only the AIO capture route.
Scroll the table horizontally to compare all columns. Keyboard users: Tab to the table, then use the arrow keys.
| ID | Exact saved question | Cohort | Attempts | Usable | Unavailable | Valid no-overview | Distinct text within question | Additional identical-text records |
|---|---|---|---|---|---|---|---|---|
| Q1 | On gistgeo.com, what is the difference between the free Gist GEO scan and the ongoing Gist GEO platform? Link to the official explanation. | Explicit lookup control | 16 | 9 | 7 | 0 | 4 | 5 |
| Q2 | Find the gistgeo.com guide comparing GA4 with AI visibility tools. Why does it say that 12 citations and six AI-referral visits do not establish a 50% click-through rate? | Explicit lookup control | 48 | 25 | 23 | 0 | 14 | 11 |
| Q3 | Find Gist’s full Halloween research report on gistgeo.com. What did it find about renting an elaborate costume for one night, and what were the collection date and answer counts for that comparison? | Explicit lookup control | 48 | 22 | 26 | 0 | 22 | 0 |
| Q4 | Find Gist’s full skincare research report on gistgeo.com. What example shows why a product being named in every answer does not mean it was recommended without reservations? | Explicit lookup control | 48 | 25 | 23 | 0 | 14 | 11 |
| Q5 | According to Gist Local on gistgeo.com, how were Los Angeles businesses selected and their websites checked, and does a high check count mean the business is recommended? | Explicit lookup control | 36 | 11 | 25 | 0 | 8 | 3 |
| Q6 | What can I learn by comparing AI referral traffic in GA4 with citations reported by an AI visibility tool, and what comparisons are misleading? | Targeted unbranded | 24 | 9 | 15 | 0 | 6 | 3 |
| Q7 | After an acquisition, how should a website explain the relationship between the parent company and a product that still uses its old name? | Targeted unbranded | 48 | 47 | 1 | 0 | 13 | 34 |
| Q8 | How should a publisher check which version of an article an AI answer actually cites when the story has been syndicated? | Targeted unbranded | 8 | 8 | 0 | 0 | 6 | 2 |
| Q9 | I write a newsletter and sell a course under my own name, but there is another creator with the same name. How can I check whether AI keeps the person, newsletter and course straight? | Targeted unbranded | 12 | 12 | 0 | 0 | 8 | 4 |
| Q10 | How should we compare AI visibility over time when we have added, removed or rewritten tracked questions? | Targeted unbranded | 48 | 47 | 1 | 0 | 19 | 28 |
| Q11 | How do I compare the real cost of AI visibility tools when they count prompts, answers and AI engines differently? | Targeted unbranded | 12 | 10 | 1 | 1 | 9 | 1 |
| Q12 | An AI visibility vendor says its agency reporting is white-label. What should I ask it to demonstrate if clients need their own login and must not see another client’s data? | Targeted unbranded | 48 | 42 | 6 | 0 | 18 | 24 |
| Q13 | Some of our AI monitoring queries failed and others returned no AI Overview. How should those results appear in our weekly visibility report? | Targeted unbranded | 24 | 17 | 7 | 0 | 7 | 10 |
| Q14 | How can I verify an AI visibility tool’s recommendation score when an answer includes both praise and reservations about a product? | Targeted unbranded | 20 | 18 | 2 | 0 | 15 | 3 |
| Q15 | I can review a small set of AI answers myself. How do I decide whether to keep using a spreadsheet or pay for an AI visibility tool? | Targeted unbranded | 12 | 8 | 4 | 0 | 5 | 3 |
| Q16 | What research has examined the advice AI assistants give about Halloween costumes beyond store recommendations? Cite sources you can verify, and say if you cannot find any. | Targeted unbranded | 16 | 14 | 2 | 0 | 10 | 4 |
| Q17 | What evidence shows whether retailers mentioned in AI shopping advice are also the sites cited as sources? Use a verifiable Halloween example if available, and say if you cannot find one. | Targeted unbranded | 12 | 9 | 3 | 0 | 6 | 3 |
| Q18 | How should a skincare company check whether its AI monitoring competitor list is missing relevant brands? Cite a real published analysis if you can verify one; otherwise say no example was found. | Targeted unbranded | 12 | 10 | 2 | 0 | 10 | 0 |
| Q19 | What should an analyst check before trusting brand counts from saved Google AI Overview answers about skincare? Cite verifiable research if available; if not, say so. | Targeted unbranded | 36 | 9 | 27 | 0 | 5 | 4 |
| Q20 | When switching AI visibility tools, what should we export so we can still audit old results after cancelling? | Targeted unbranded | 12 | 12 | 0 | 0 | 6 | 6 |
Four complete selected saved texts
These are unedited generated outputs for evidence inspection, not endorsed advice or a fresh factual audit. Original wording—including an incomplete bullet—is preserved. The saved citation lists do not establish verified claim-to-source support.
AIO-Q07-E8 · Targeted unbranded · 2026-10-29T19:15:13.633Z
Exact question: After an acquisition, how should a website explain the relationship between the parent company and a product that still uses its old name?
Google AI Overviews capture route · saved 2026-10-29T19:15:13.633Z
After an acquisition, a website should explain the parent-product relationship using an endorsed brand strategy, where the product keeps its familiar name while a clear, concise descriptor or lockup links it to the parent company. Core Ways to Explain the Relationship - Use an Endorsed Sub-Brand Line: Add a small sign-off or sub-line beneath the product logo, in the header, or in the hero section (e.g., "[Product Name], a [Parent Company] company" or "[Product Name], backed by [Parent Company]"). - Draft a Dedicated "About" / FAQ Section: Write a short, transparent paragraph on the About Us, FAQ, or announcement page explaining why the acquisition happened and what it means for users (e.g., "On [Date], [Product] was acquired by [Parent Company]. While our name and core experience remain the same, we now have access to [Parent Company’s] resources to build better features faster"). - Update the Website Footer: Place the parent company’s legal name and copyright in the footer to signal institutional ownership without cluttering the main marketing message (e.g., © 2026 [Parent Company]. All rights reserved. [Product Name] is a registered trademark of [Parent Company].). - Employ a Transitional Banner or Modal: If the acquisition is fresh, use a non-intrusive top announcement bar or banner pointing to a press release or FAQ page to reassure existing customers that service continuity is uninterrupted. Best Practices for Messaging - Reassure Continuity: Emphasize what is not changing (login, pricing, support, product name) before introducing the parent company. - Highlight the Value Add: Explain how the parent company helps the product (e.g., stronger security, faster engineering, or broader support) rather than just stating corporate ownership. - Keep Tone Aligned: Ensure the voice remains true to the product's established community while adopting the stability of the parent brand. If you'd like, share: - The product name and parent company - Whether you are trying to calm existing users or signal enterprise credibility I can help you draft the exact copy for your website banner, About page, or footer.
Saved citation list — not independently verified support:
- https://embarkagency.com.au/insights/brand-architecture-examples
- https://studiosunup.com/blog/acquisition-branding/
- https://www.marrow.red/insights/brand-strategy/brand-architecture-after-a-merger-or-acquisition-retain-endorse-or-integrate/
- https://hekatestrategies.com/post-acquisition-brand-architecture/
- https://usercentrics.com/guides/website-disclaimers/copyright-footer/
AIO-Q07-E32 · Targeted unbranded · 2026-10-29T19:24:32.855Z
Exact question: After an acquisition, how should a website explain the relationship between the parent company and a product that still uses its old name?
Google AI Overviews capture route · saved 2026-10-29T19:24:32.855Z
After an acquisition, a website should explain the parent-product relationship using an endorsed brand strategy, where the product keeps its familiar name while a clear, concise descriptor or lockup links it to the parent company. Core Ways to Explain the Relationship - Use an Endorsed Sub-Brand Line: Add a small sign-off or sub-line beneath the product logo, in the header, or in the hero section (e.g., "[Product Name], a [Parent Company] company" or "[Product Name], backed by [Parent Company]"). - Draft a Dedicated "About" / FAQ Section: Write a short, transparent paragraph on the About Us, FAQ, or announcement page explaining why the acquisition happened and what it means for users (e.g., "On [Date], [Product] was acquired by [Parent Company]. While our name and core experience remain the same, we now have access to [Parent Company’s] resources to build better features faster"). - Update the Website Footer: Place the parent company’s legal name and copyright in the footer to signal institutional ownership without cluttering the main marketing message (e.g., © 2026 [Parent Company]. All rights reserved. [Product Name] is a registered trademark of [Parent Company].). - Employ a Transitional Banner or Modal: If the acquisition is fresh, use a non-intrusive top announcement bar or banner pointing to a press release or FAQ page to reassure existing customers that service continuity is uninterrupted. Best Practices for Messaging - Reassure Continuity: Emphasize what is not changing (login, pricing, support, product name) before introducing the parent company. - Highlight the Value Add: Explain how the parent company helps the product (e.g., stronger security, faster engineering, or broader support) rather than just stating corporate ownership. - Keep Tone Aligned: Ensure the voice remains true to the product's established community while adopting the stability of the parent brand. If you'd like, share: - The product name and parent company - Whether you are trying to calm existing users or signal enterprise credibility I can help you draft the exact copy for your website banner, About page, or footer.
Saved citation list — not independently verified support:
- https://embarkagency.com.au/insights/brand-architecture-examples
- https://studiosunup.com/blog/acquisition-branding/
- https://www.marrow.red/insights/brand-strategy/brand-architecture-after-a-merger-or-acquisition-retain-endorse-or-integrate/
- https://hekatestrategies.com/post-acquisition-brand-architecture/
- https://usercentrics.com/guides/website-disclaimers/copyright-footer/
AIO-Q03-E3 · Explicit lookup control · 2026-10-29T19:12:13.597Z
Exact question: Find Gist’s full Halloween research report on gistgeo.com. What did it find about renting an elaborate costume for one night, and what were the collection date and answer counts for that comparison?
Google AI Overviews capture route · saved 2026-10-29T19:12:13.597Z
According to Gist's Halloween research report found on Gist GEO, when evaluating a prompt that compared renting an elaborate Halloween costume for one night versus buying one, the AI models frequently favored renting due to the one-night premise. Specifically, the report highlighted the following data points for this specific comparison: - Collection Date: All responses were collected on October 9, 2026. - ChatGPT Answer Counts: 48 saved answers were analyzed. Out of these: - 48 out of 48 favored rental under the one-night premise. - 27 out of 48 also mentioned secondhand options. - 1 out of 48 mentioned borrowing. - Perplexity Answer Counts: 16 saved answers were analyzed for this section. If you are interested, I can provide more details on how general retailers versus costume specialists performed across the other 29 questions in Gist's report, or dive deeper into their AI search optimization (GEO) methodology. How would you like to proceed?
Saved citation list — not independently verified support:
AIO-Q03-E4 · Explicit lookup control · 2026-10-29T19:14:46.290Z
Exact question: Find Gist’s full Halloween research report on gistgeo.com. What did it find about renting an elaborate costume for one night, and what were the collection date and answer counts for that comparison?
Google AI Overviews capture route · saved 2026-10-29T19:14:46.290Z
In its October 2026 report, Halloween shopping advice beyond the complete costume, Gist analyzed how AI engines evaluate the decision between renting an elaborate costume for one night versus purchasing it. Key Findings on Costume Rental The report found that under the specific premise of needing an elaborate costume for just one night, the AI answers favored rental conditionally, subject to underlying costs and rental terms. However, the models still highlighted circumstances where buying would make more sense (such as repeated future wear, heavy customization needs, or strict rental deposit and return obligations). Additionally, the models looked past a strict rent-versus-buy binary by introducing alternative acquisition channels: - Secondhand options were frequently brought up in the answers as a viable alternative. - was suggested by the models as another method to acquire the look without buying. Collection Date & Answer Counts The empirical details for this specific rental comparison (Question ID: HALLOWEEN-10-1) include: - Collection Date: October 9, 2026 - Total Answer Counts: 64 successful answers in total for this comparison, broken down as: - ChatGPT (OpenAI API): 48 saved answers - Perplexity (Sonar API): 16 saved answers - Behavior Breakdowns within the Counts: - Favored rental conditionally: 48/48 ChatGPT answers and 16/16 Perplexity answers. - Mentioned secondhand: 27/48 ChatGPT answers and 12/16 Perplexity answers. - Mentioned borrowing: 1/48 ChatGPT answer and 10/16 Perplexity answers. If you would like to explore more specifics, I can detail what Gist discovered regarding long-term reuse calculations or how budget constraints changed which retailers the AI recommended. Which section should we look at next?
Saved citation list — not independently verified support:
No URLs are present in this saved citation-list field.