---
title: "59% of ChatGPT Citations Disappeared in a Day. Don’t Hire or Fire Over a Screenshot."
description: "Citation turnover makes one ChatGPT screenshot weak evidence. Here’s the weekly record to demand before hiring or firing an AI search provider."
url: "https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/804497c9-8d45-4026-84a5-6149041b8c9d.png"
published: "2026-11-09T06:24:40.066Z"
modified: "2026-11-09T06:24:40.130Z"
---

November 9, 2026 · 9 min read

# 59% of ChatGPT Citations Disappeared in a Day. Don’t Hire or Fire Over a Screenshot.

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![A hand holds an instant photo of a rainbow over a field, lined up against the same field now grey and rainy, showing that a snapshot captures only a passing moment.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/804497c9-8d45-4026-84a5-6149041b8c9d.png)

In this article

1. [A citation is not a recommendation](#a-citation-is-not-a-recommendation)
2. [What the citation-turnover study measured](#what-the-citation-turnover-study-measured)
3. [What the numbers cannot tell you](#what-the-numbers-cannot-tell-you)
4. [Replace the screenshot with a weekly evidence record](#replace-the-screenshot-with-a-weekly-evidence-record)
5. [Require a work log, not just a better chart](#require-a-work-log-not-just-a-better-chart)

[59% of ChatGPT citations present on day one disappeared by day two](https://ahrefs.com/blog/how-stable-are-ai-citations/). That makes one missing citation a weak reason to fire your AI search provider—and one favorable screenshot a weak reason to hire one.

The same study tracked identical questions across five tools for [20 days](https://ahrefs.com/blog/how-stable-are-ai-citations/). About [75% of lost citations returned during that window](https://ahrefs.com/blog/how-stable-are-ai-citations/). Disappearance was common. So was coming back.

**A screenshot records an appearance, not durable visibility.** Before I credit a provider for a win or blame them for a loss, I want to know how often the brand appears, what the answer actually says, and whether any of it connects to business results.

## A citation is not a recommendation

A citation is a footnote: your page appears in the answer’s source list. That is different from the tool naming your brand in the text, and different again from recommending you as the pick.

One [analysis of citations and brand mentions](https://soedja.com/blog/ai-citation-stability-geo-2026) found that many cited brands never appeared in the answer text:

| Tool         | Citations without a brand mention in the answer               |
| ------------ | ------------------------------------------------------------- |
| AI Mode      | [84%](https://soedja.com/blog/ai-citation-stability-geo-2026) |
| AI Overviews | [77%](https://soedja.com/blog/ai-citation-stability-geo-2026) |
| ChatGPT      | [54%](https://soedja.com/blog/ai-citation-stability-geo-2026) |

These figures describe the gap between being sourced and being named in that analysis. They do not measure how often a buyer chooses the brand, and they are not a forecast for your account.

I used to tell clients a new link meant we were winning. I was wrong. A _ghost citation_, where you supply the source and a competitor gets the recommendation, can leave a visibility report looking healthier than the answer a buyer reads.

Operators who audit this work separate [being cited, being named, and being recommended](https://www.onely.com/blog/why-ai-chatbots-cite-your-content-but-dont-recommend-your-brand-by-name/). I want the same separation in a provider’s report:

- **Cited:** your page appears as a source.
- **Named:** your brand appears in the answer text.
- **Recommended:** the answer presents your brand as a choice.

Count those separately, or you can pay for footnotes and think you bought customers.

Repeated checks also change the picture. One analysis found only [41% domain overlap when repeating the same ChatGPT prompt on the same day](https://spb.trendi.ai/tag/ai-citation-changes/). A separate series of [10,826 daily runs](https://spb.trendi.ai/tag/ai-citation-changes/) found that [70% of brands that ever appeared showed up on fewer than 20% of observed days](https://spb.trendi.ai/tag/ai-citation-changes/). Only [6% appeared on at least 80% of days](https://spb.trendi.ai/tag/ai-citation-changes/).

Those are separate observations, not interchangeable measures. Their practical message is consistent: one check tells me very little about frequency.

This cuts both ways. A screenshot showing you present today does not prove durable visibility. A screenshot showing you absent tomorrow does not prove you lost it. If an agency sends either alone as evidence, it is selling weather as climate.

## What the citation-turnover study measured

Ahrefs asked [7,831 identical questions daily for 20 days](https://ahrefs.com/blog/how-stable-are-ai-citations/), split across five tools. It followed roughly [37,500 websites cited on day one](https://ahrefs.com/blog/how-stable-are-ai-citations/) to see whether they remained in subsequent answers.

Same question, same tool, repeated daily. That design makes answer churn visible instead of letting one favorable response stand in for a trend.

| Tool         | Questions tracked                                             | Day-one cited sites absent on day two                       |
| ------------ | ------------------------------------------------------------- | ----------------------------------------------------------- |
| AI Overviews | [563](https://ahrefs.com/blog/how-stable-are-ai-citations/)   | [50%](https://ahrefs.com/blog/how-stable-are-ai-citations/) |
| AI Mode      | [1,996](https://ahrefs.com/blog/how-stable-are-ai-citations/) | [58%](https://ahrefs.com/blog/how-stable-are-ai-citations/) |
| ChatGPT      | [3,173](https://ahrefs.com/blog/how-stable-are-ai-citations/) | [59%](https://ahrefs.com/blog/how-stable-are-ai-citations/) |
| Gemini       | [1,276](https://ahrefs.com/blog/how-stable-are-ai-citations/) | [81%](https://ahrefs.com/blog/how-stable-are-ai-citations/) |
| Copilot      | [823](https://ahrefs.com/blog/how-stable-are-ai-citations/)   | [83%](https://ahrefs.com/blog/how-stable-are-ai-citations/) |

Disappearance is only half the story. The study also tracked returns and persistence:

| Observation                                                                     | Finding                                                            |
| ------------------------------------------------------------------------------- | ------------------------------------------------------------------ |
| Lost citations returning during the study                                       | [About 75%](https://ahrefs.com/blog/how-stable-are-ai-citations/)  |
| Lost citations returning within three days                                      | [About half](https://ahrefs.com/blog/how-stable-are-ai-citations/) |
| Lost citations returning after a single missing day on AI Overviews and ChatGPT | [About 40%](https://ahrefs.com/blog/how-stable-are-ai-citations/)  |
| Citations lasting the entire study on Google tools and ChatGPT                  | [4.0–6.8%](https://ahrefs.com/blog/how-stable-are-ai-citations/)   |

Persistence mattered too. Citations seen on only one day stayed gone [34–41% of the time](https://ahrefs.com/blog/how-stable-are-ai-citations/), while those present for [8–14 days returned 96–99% of the time](https://ahrefs.com/blog/how-stable-are-ai-citations/).

**Short appearances flicker. Longer appearances still flicker, but usually come back.** That is a better starting point for reading a visibility report than assuming every disappearance reflects something a provider broke.

## What the numbers cannot tell you

The disappearance rates above describe those questions, in that wording, over that observation window. They are not your disappearance rate.

Change the wording, change the location, or ask a bottom-funnel question with strong competitors instead of an informational question with many possible sources, and the result can differ. The study does not tell a plumber in Austin or a SaaS vendor selling to finance teams what share of their buyer prompts will cite them tomorrow.

The limit is straightforward: **the data establishes that churn is common; it does not forecast your account.** Anyone applying the table directly to your business is going beyond what it supports.

It also does not establish what caused a citation to return. When roughly [three in four lost citations reappear during the observation window](https://ahrefs.com/blog/how-stable-are-ai-citations/), seeing your link return after a page edit is weak evidence that the edit worked. The timing alone cannot distinguish an improvement from ordinary churn.

Ahrefs makes the same practical point: do not credit a page change for a single return, and do not judge visibility from one check. I approach it the way I approach ad copy tests. I want fixed prompts measured before and after, plus a holdout set left untouched. Otherwise, the chart can lie in the friendliest possible way.

My rule: a returning citation is not an intervention win until repeated measurement supports that interpretation. More frequent appearances give me a pattern to investigate. They do not, by themselves, explain its cause.

An agency that claims every favorable fluctuation as progress has no credible way to explain the unfavorable ones.

## Replace the screenshot with a weekly evidence record

If I am paying someone to manage how my brand appears in ChatGPT, Gemini, and Perplexity, I want a ledger. Screenshots can illustrate it. They cannot replace it.

A useful [AI visibility report fixes the prompt universe, samples repeatedly, and tracks presence, prominence, citations, and competitive share over time](https://mavlers.gush.site/blog/ai-visibility-report-guide). I want commercial results alongside that record, not folded into a visibility score.

The setup is straightforward:

1. **Choose buyer questions.** Cover the situations that precede a purchase.
2. **Freeze the wording.** Otherwise, a changed question can masquerade as changed performance.
3. **Repeat observations across the week.** Log the prompt, tool, date, and result for each run.
4. **Separate citations, mentions, and recommendations.** They answer different questions.
5. **Keep competitors and commercial results visible.** Neither belongs inside a blended visibility percentage.

The prompt set should cover these jobs:

| Prompt type   | Example                                          |
| ------------- | ------------------------------------------------ |
| Category      | “Best CRM for an agency”                         |
| Comparison    | “X vs Y for invoicing”                           |
| Problem-first | “How do I reduce no-shows for a dental clinic?”  |
| Brand-direct  | “Is \[your brand] good for…?”                    |
| Bottom-funnel | “\[Your brand] pricing vs \[competitor] pricing” |

I want daily observations if the report uses a daily presence metric. Ahrefs recommends [counting how many of the last seven days you were cited](https://ahrefs.com/blog/how-stable-are-ai-citations/), prompt by prompt. If the provider repeats a prompt within a day, the underlying log should preserve those runs rather than letting the most favorable answer become the day’s entire story.

The following is an **illustrative reporting layout, not measured results**. Each fraction means days observed with that outcome out of seven; the accompanying log needs the actual dates.

| Buyer prompt                              | ChatGPT                               | Gemini                                | Perplexity                            |
| ----------------------------------------- | ------------------------------------- | ------------------------------------- | ------------------------------------- |
| Best invoicing tool for an agency         | Cited 5/7; named 3/7; recommended 1/7 | Cited 2/7; named 1/7; recommended 0/7 | Cited 4/7; named 2/7; recommended 1/7 |
| \[Your brand] vs \[competitor] for salons | Cited 6/7; named 6/7; recommended 2/7 | Cited 4/7; named 3/7; recommended 1/7 | Cited 3/7; named 3/7; recommended 0/7 |

This format makes an uncomfortable result obvious: a brand can be cited regularly and rarely recommended. A single visibility score can hide that gap.

At the weekly review, I would read across each prompt before looking at an overall total. Is the brand supplying evidence without becoming a suggested choice? Is a competitor getting the recommendation? Those are different problems from not appearing at all. I want the provider to name the gap, not bury it under a rising citation count.

I also want qualified inquiries and attributable revenue reported for the same period, separately from the visibility counts. If citations rise and pipeline does not, the report has identified a gap worth investigating. It has not proved that visibility caused revenue—or that visibility can be ruled out as a constraint.

**Keep the money beside the visibility, not inside it.**

## Require a work log, not just a better chart

Repeated measurement tells me what happened. I still need to know what the provider did.

Some of the work associated with citations is unglamorous and structural. A study comparing [48,527 cited pages with uncited pages from the same domains](https://www.sistrix.com/blog/citation-worthiness-study/) found these associations:

| Page feature             | Reported citation association                                                  |
| ------------------------ | ------------------------------------------------------------------------------ |
| Many structural features | [2.1× as often cited](https://www.sistrix.com/blog/citation-worthiness-study/) |
| Table of contents        | [1.7×](https://www.sistrix.com/blog/citation-worthiness-study/)                |
| Article schema           | [1.6×](https://www.sistrix.com/blog/citation-worthiness-study/)                |
| FAQ schema               | [1.5×](https://www.sistrix.com/blog/citation-worthiness-study/)                |

Those are associations, not a promise that adding markup produces the same lift. The same study found [no advantage for llms.txt: 11.7% of cited domains used one, versus 14.0% of all domains](https://www.sistrix.com/blog/citation-worthiness-study/).

So when a proposal leads with a text file and a blog calendar, I pass. I want clean product data, tight category pages, quotable definitions, comparison tables a model can lift intact, and third-party corroboration. Then I want the work listed by URL and date, next to repeated measurements—not presented as a guaranteed route to citations.

Before paying, I ask for:

- **A dated work log:** what changed, where, and when.
- **A repeatable measurement record:** fixed prompts, named tools, observation dates, and consistent sampling.
- **Separate outcomes:** citations, mentions, recommendations, competitor presence, and qualified inquiries.
- **Written limits:** what the provider will not claim from the evidence.

I want those records to connect without pretending they prove causation. If the provider edits a comparison page, show me the dated change and the relevant prompt history before and after it. Keep the untouched prompts visible too. That gives us something to question together, rather than a success story assembled after the answer happens to improve.

That is the [weekly evidence record I explain in detail here](https://groas.com/post/your-ai-visibility-measurement-questions). A provider willing to expose the underlying observations gives me something to evaluate. A provider unwilling to show frequency has not given me evidence of durable visibility.

The difference is accountability, not prettier reporting.

### No guaranteed citations. No victory laps over one answer.

A promise to put your brand in every ChatGPT answer ignores the instability these observations expose. A favorable answer today cannot guarantee the same source list tomorrow.

That is why I buy ongoing execution and accountable measurement, not a promised outcome on a single prompt. At [groas](https://groas.com/), the model I trust pairs continuous execution with a named strategist who owns the result and a log of changes. Catalog fixes, comparison content, third-party corroboration, then the same prompts measured again: work I can inspect and results I can question.

**The fee should buy accountable work, not a guarantee that the model will behave.**

If your brand vanished today, do not fire anyone on that evidence alone. Freeze your buyer questions, repeat the observations, and ask for presence by tool with dates, mentions and recommendations in separate columns, and pipeline beside them. The setup questions are covered in [_Does ChatGPT Recommend Your Business? Seven Questions About Measuring AI Visibility_](https://groas.com/post/reader-q-a-how-do-i-actually-measure-whe).

Hire the provider who will show you that ledger every week, limits included. Fire the one who keeps sending screenshots.

## Frequently Asked Questions

### Can I hire or fire my AI search provider based on a single ChatGPT screenshot?

No. An Ahrefs study of 7,831 identical questions asked daily for 20 days found that 59% of ChatGPT citations present on day one were gone by day two, and about 75% of lost citations returned during the window. One screenshot records an appearance, not durable visibility, so it is weak evidence for either hiring or firing.

### What is the difference between being cited, named, and recommended in an AI answer?

Cited means your page appears as a source in the answer's list. Named means your brand appears in the answer text itself. Recommended means the answer presents your brand as a choice. These should be counted separately, because an analysis found many cited brands never appeared in the answer text: 84% on AI Mode, 77% on AI Overviews, and 54% on ChatGPT.

### What is a ghost citation in AI visibility reporting?

A ghost citation is when your page supplies the source for an AI answer but a competitor gets the recommendation in the text. The visibility report looks healthier than the answer a buyer actually reads. That is why citations, mentions, and recommendations need separate counts rather than one blended score.

### How did the Ahrefs citation stability study actually work?

Ahrefs asked 7,831 identical questions daily for 20 days, split across five AI tools, and followed roughly 37,500 websites cited on day one to see whether they stayed cited. Because the same question was repeated daily on the same tool, the study makes answer churn visible instead of letting one favorable response stand in for a trend.

### Do these AI citation disappearance rates apply to my business?

No. The rates describe those specific questions, in that wording, over that observation window. Change the wording, the location, or whether the question is informational versus bottom-funnel, and the result can differ. The data establishes that citation churn is common; it does not forecast what share of your buyer prompts will cite you.

### If my citation came back right after I edited a page, did the edit work?

Not necessarily. About three in four lost citations reappear during the observation window anyway, so the timing of a single return cannot distinguish an improvement from ordinary churn. A returning citation counts as an intervention win only when repeated measurement, such as fixed prompts measured before and after with a holdout set, supports that interpretation.

### What should an AI visibility report include instead of screenshots?

A ledger that freezes the wording of buyer questions, repeats observations across the week, and logs the prompt, tool, date, and result for each run. It should separate citations, mentions, and recommendations, keep competitor presence visible, and report qualified inquiries and attributable revenue alongside the visibility counts, not blended into one score.

### Does adding llms.txt help my site get cited by AI tools?

The Sistrix citation-worthiness study found no advantage for llms.txt: 11.7% of cited domains used one, versus 14.0% of all domains. The same study did find associations with structural features, such as a table of contents (1.7 times as often cited), article schema (1.6 times), and FAQ schema (1.5 times).

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: free to ask, $49 a month flat with every feature included, and autopilot whenever you’re ready.

[Start Free](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to (free to read, $49 a month to make changes, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#webpage","url":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove","name":"59% of ChatGPT Citations Disappeared in a Day. Don’t Hire or Fire Over a Screenshot.","description":"Citation turnover makes one ChatGPT screenshot weak evidence. Here’s the weekly record to demand before hiring or firing an AI search provider.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/804497c9-8d45-4026-84a5-6149041b8c9d.png"},"breadcrumb":{"@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#breadcrumb"},"dateModified":"2026-11-09T06:24:40.130Z","mainEntity":{"@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"59% of ChatGPT Citations Disappeared in a Day. Don’t Hire or Fire Over a Screenshot.","item":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#article","url":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove","mainEntityOfPage":{"@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"59% of ChatGPT Citations Disappeared in a Day. Don’t Hire or Fire Over a Screenshot.","description":"Citation turnover makes one ChatGPT screenshot weak evidence. Here’s the weekly record to demand before hiring or firing an AI search provider.","datePublished":"2026-11-09T06:24:40.066Z","dateModified":"2026-11-09T06:24:40.130Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/804497c9-8d45-4026-84a5-6149041b8c9d.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"citations, chatgpt, cited, day, screenshot, visibility, brand, citation","wordCount":1875,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/your-brand-vanished-from-chatgpt-today-that-doesn-t-prove#faq","mainEntity":[{"@type":"Question","name":"Can I hire or fire my AI search provider based on a single ChatGPT screenshot?","acceptedAnswer":{"@type":"Answer","text":"No. An Ahrefs study of 7,831 identical questions asked daily for 20 days found that 59% of ChatGPT citations present on day one were gone by day two, and about 75% of lost citations returned during the window. One screenshot records an appearance, not durable visibility, so it is weak evidence for either hiring or firing."}},{"@type":"Question","name":"What is the difference between being cited, named, and recommended in an AI answer?","acceptedAnswer":{"@type":"Answer","text":"Cited means your page appears as a source in the answer's list. Named means your brand appears in the answer text itself. Recommended means the answer presents your brand as a choice. These should be counted separately, because an analysis found many cited brands never appeared in the answer text: 84% on AI Mode, 77% on AI Overviews, and 54% on ChatGPT."}},{"@type":"Question","name":"What is a ghost citation in AI visibility reporting?","acceptedAnswer":{"@type":"Answer","text":"A ghost citation is when your page supplies the source for an AI answer but a competitor gets the recommendation in the text. The visibility report looks healthier than the answer a buyer actually reads. That is why citations, mentions, and recommendations need separate counts rather than one blended score."}},{"@type":"Question","name":"How did the Ahrefs citation stability study actually work?","acceptedAnswer":{"@type":"Answer","text":"Ahrefs asked 7,831 identical questions daily for 20 days, split across five AI tools, and followed roughly 37,500 websites cited on day one to see whether they stayed cited. Because the same question was repeated daily on the same tool, the study makes answer churn visible instead of letting one favorable response stand in for a trend."}},{"@type":"Question","name":"Do these AI citation disappearance rates apply to my business?","acceptedAnswer":{"@type":"Answer","text":"No. The rates describe those specific questions, in that wording, over that observation window. Change the wording, the location, or whether the question is informational versus bottom-funnel, and the result can differ. The data establishes that citation churn is common; it does not forecast what share of your buyer prompts will cite you."}},{"@type":"Question","name":"If my citation came back right after I edited a page, did the edit work?","acceptedAnswer":{"@type":"Answer","text":"Not necessarily. About three in four lost citations reappear during the observation window anyway, so the timing of a single return cannot distinguish an improvement from ordinary churn. A returning citation counts as an intervention win only when repeated measurement, such as fixed prompts measured before and after with a holdout set, supports that interpretation."}},{"@type":"Question","name":"What should an AI visibility report include instead of screenshots?","acceptedAnswer":{"@type":"Answer","text":"A ledger that freezes the wording of buyer questions, repeats observations across the week, and logs the prompt, tool, date, and result for each run. It should separate citations, mentions, and recommendations, keep competitor presence visible, and report qualified inquiries and attributable revenue alongside the visibility counts, not blended into one score."}},{"@type":"Question","name":"Does adding llms.txt help my site get cited by AI tools?","acceptedAnswer":{"@type":"Answer","text":"The Sistrix citation-worthiness study found no advantage for llms.txt: 11.7% of cited domains used one, versus 14.0% of all domains. The same study did find associations with structural features, such as a table of contents (1.7 times as often cited), article schema (1.6 times), and FAQ schema (1.5 times)."}}]}]}
```
