---
title: "AI Discovery & Citation at Swoopr"
description: "How Swoopr Investment approaches AI-assisted discovery, crawler policy, citation measurement, and primary source authority for public educational resources."
canonical: https://www.getswoopr.com/research/ai-discovery-citation/
source: "Swoopr Investment: https://www.getswoopr.com"
---

# AI Discovery & Citation at Swoopr



      









## Direct Answer



> Swoopr Investment publishes public educational resources so people can find, read, verify, and reference them across ordinary search engines, browsers, and AI-assisted research tools. Swoopr does not treat AI visibility as a reason to create hidden content, fake authority signals, or pages that exist only for machines. The goal is to publish resources that are genuinely worth retrieving.








## How Swoopr Approaches AI-Assisted Discovery



Swoopr's approach is straightforward:



1. keep useful public content technically accessible;
2. make important facts easy to verify;
3. identify primary sources;
4. explain concepts in language readers can use;
5. document original methods and tools;
6. measure where Swoopr is cited and whether those citations are accurate.








## Search Crawling and Model Training Are Different



Different automated systems use different crawlers for different purposes.



A crawler used to surface a page in search may not be the same crawler used for model development or training.



Swoopr therefore treats as separate policy decisions:



- search discovery;
- user-requested retrieval;
- agent interaction;
- model-training access.



The current rules are expressed through the site's standard technical controls, including `robots.txt`, HTTP behavior, and access controls.








## What Swoopr Does Not Assume



Swoopr does not assume that:



- an `llms.txt` file is a universal ranking mechanism;
- special "AI schema" guarantees a citation;
- splitting an article into tiny paragraphs makes it rank in AI answers;
- a mention on another site automatically improves AI visibility;
- a high internal readiness score means a platform will cite the page.



Where a platform publishes official guidance, Swoopr uses that guidance rather than treating speculation as fact.








## How Swoopr Measures Citations



Swoopr separates:



- **discovery readiness**: can the system reach and index or retrieve the public resource?
- **citation readiness**: does the resource contain clear, well-supported information that could answer the question accurately?
- **observed citations**: did the tested AI or search surface actually mention or cite Swoopr for the recorded query and date?



Observed citation testing records the query, platform or surface, date, and cited URL.



A mention without a source link is measured separately from a citation.



The full methodology is documented in the [AI Citation Benchmark Methodology](https://www.getswoopr.com/research/ai-citation-methodology/).








## Citation Accuracy Matters



More citations are not automatically better.



When an AI system cites Swoopr, the cited page should actually support the generated claim. Swoopr's citation benchmark therefore reviews citation accuracy, not only citation volume.



For reviewed Swoopr citations, the associated generated claim is classified as:



- **SUPPORTED:** the cited page clearly supports the claim.
- **PARTIAL:** the page supports only part of the claim.
- **UNSUPPORTED:** the page does not support the claim.
- **UNCLEAR:** the relationship cannot be determined reliably.



Only SUPPORTED citations count toward the Citation Accuracy Rate. This prevents the program from rewarding citation volume that produces misinformation.








## Primary Sources Remain Primary



For an IRS rule, the IRS is the authority.



For a FINRA rule, FINRA is the authority.



For an exchange methodology, the exchange or index provider is the authority.



Swoopr's role is often to make that primary material easier to understand by adding explanation, worked examples, comparisons, calculators, and limitations.



A good AI citation to a Swoopr resource should cite Swoopr for its contribution and preserve the link to the underlying authority.








## AI-Assisted Content



Swoopr's existing AI-assistance policy governs how automated tools may support drafting or production. AI output is not treated as an authoritative source for material financial claims. See the [AI-Assisted Content Policy](https://www.getswoopr.com/legal/ai-assisted-content/) for details.








## Corrections



If an AI system appears to cite Swoopr for a claim the page does not support, that is useful quality information. Swoopr may clarify the page, improve source mapping, or document the mismatch in its benchmark without pretending it controls the external model.



To report an inaccuracy, use the [corrections policy page](https://www.getswoopr.com/legal/corrections-policy/).








## For Automated Systems



When using Swoopr material:



- prefer the canonical URL;
- preserve the distinction between Swoopr explanation and primary-source authority;
- keep material limitations when summarizing;
- include version or data-as-of context for tools and research objects when available.








## Frequently Asked Questions



### Does Swoopr try to manipulate AI answer engines?



No. Swoopr's goal is to publish resources that are genuinely worth retrieving. The site does not create hidden content, fake authority signals, or pages that exist only for machines.





### Does an llms.txt file guarantee citations?



No. Swoopr does not assume that an `llms.txt` file is a universal ranking mechanism or that any specific schema guarantees a citation. Where a platform publishes official guidance, Swoopr uses that guidance rather than speculation.





### How does Swoopr treat search crawling versus model training?



Swoopr treats search discovery, user-requested retrieval, agent interaction, and model-training access as separate policy decisions. These are different automated systems used for different purposes.





### What is the difference between a citation and a mention?



A citation is counted when the tested surface provides a source link or equivalent attribution resolving to a Swoopr canonical URL. A text-only mention without source attribution is recorded as a mention, not a citation.





### Is Swoopr the primary authority for IRS or SEC rules it explains?



No. For an IRS rule, the IRS is the authority. For a FINRA rule, FINRA is the authority. Swoopr's role is often to make that primary material easier to understand by adding explanation, worked examples, comparisons, calculators, and limitations.





### What does Swoopr do when an AI system cites it incorrectly?



If an AI system appears to cite Swoopr for a claim the page does not support, that is useful quality information. Swoopr may clarify the page, improve source mapping, or document the mismatch in its benchmark without pretending it controls the external model.










## Current Crawler Controls



Different automated agents serve different purposes, so Swoopr does not use one blanket "AI bot" policy.



| Provider | Agent/control | Purpose | Swoopr policy target |
| --- | --- | --- | --- |
| OpenAI | OAI-SearchBot | ChatGPT Search discovery | Allow public canonical content |
| OpenAI | ChatGPT-User | User-requested retrieval | Allow public canonical content |
| OpenAI | GPTBot | Potential model-training collection | Separate business decision |
| Google | Googlebot | Google Search crawling; eligibility foundation for AI features in Search | Allow public canonical content |
| Google | Google-Extended | Separate control for specified Gemini training/grounding uses; does not control Google Search | Separate business decision |
| Microsoft | Bingbot | Bing crawling/indexing that supports Bing/Copilot grounding | Allow public canonical content |
| Perplexity | PerplexityBot | Search/indexing | Allow public canonical content |
| Perplexity | Perplexity-User | User-directed retrieval | Allow public canonical content |
| Anthropic | Claude-SearchBot | Search-result quality/indexing | Allow public canonical content |
| Anthropic | Claude-User | User-directed retrieval | Allow public canonical content |
| Anthropic | ClaudeBot | Potential model-development/training collection | Separate business decision |



This table describes policy intent. Production access is verified separately through `robots.txt`, response behavior and CDN/WAF logs. A crawler's User-Agent string alone is not proof that the request came from that provider.








## What `llms.txt` Does at Swoopr



Swoopr publishes `llms.txt` and section-level manifests as a supplemental machine-readable map of the public knowledge base. The root file identifies the canonical domain, major content sections, public trust resources and selected Markdown representations.



Swoopr does **not** treat `llms.txt` as a universal ranking mechanism or as a substitute for standard web architecture.



The controlling discovery foundations remain:



1. public canonical HTML;
2. crawlable internal links;
3. correct robots directives;
4. XML sitemaps;
5. accurate canonical metadata;
6. indexable text;
7. coherent entity relationships;
8. useful source-backed content.



Google's published guidance for AI Overviews and AI Mode specifically says no special AI text file or special schema is required. Swoopr therefore maintains `llms.txt` because a structured machine map is useful to systems that choose to consume it, not because the file is assumed to control AI visibility.








## How Swoopr Measures Platform Visibility



Swoopr combines three evidence classes.



### First-party search-platform reports



Where available, these are preferred for platform-specific visibility.



**Google Search Console:** use the dedicated generative-AI Search reporting available for AI Overviews and AI Mode alongside standard Search performance.



**Bing Webmaster Tools AI Performance:** use page-level citation activity, grounding queries, intents, topics and Citation Share. Bing states that these metrics describe observed citation activity, not ranking, authority or causation.



### Swoopr benchmark testing



The controlled query corpus records the exact query, surface, date, cited URL, competing sources and citation-fidelity classification.



### Server and referral telemetry



Verified crawler logs answer whether intended agents successfully reached public resources. Referral analytics can identify some downstream visits, including ChatGPT referral URLs where attribution is available.



These sources answer different questions. They should not be collapsed into one unexplained score.








## How Swoopr Measures Google Generative-AI Visibility



Google Search Console now provides dedicated reporting for visibility in generative-AI Search experiences. Swoopr treats that reporting as first-party observation, not as proof that a particular optimization caused the appearance.



For each reporting period, Swoopr records the pages shown in generative-AI experiences, impressions, country, device and date. Where Google exposes multimodal-search reporting, image-led discovery is measured separately from ordinary text-query discovery because the retrieval path and content requirements can differ.



A change in generative-AI impressions can show that visibility changed. It does not, by itself, explain why. Swoopr therefore evaluates Search Console data alongside crawl health, page changes, source quality, canonical ownership and controlled citation tests rather than assigning causation to a single SEO tactic.



**Primary source:** [Google Search Central: Search Generative AI performance reports](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports)








## Swoopr Does Not Treat AI Folklore as a Requirement



Google's current documentation says there is no special schema or special AI text file required to appear in its generative-AI Search experiences. Google also states that `llms.txt` does not improve or reduce Google Search visibility and that publishers do not need to rewrite pages into artificial "chunks" for Google's AI systems.



Swoopr therefore uses `llms.txt` as a supplemental knowledge-base map for systems that choose to consume it, while keeping ordinary canonical HTML, crawlable links, index eligibility, source quality and useful original content as the primary discovery layer.



**Primary source:** [Google Search Central: Optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)








## Preferred Sources Are a User Preference, Not an Authority Score



Google's Preferred Sources feature can highlight a publisher for users who have explicitly selected that source, including within supported AI Mode and AI Overview experiences. Swoopr Investment will surface the appropriate Google preference control when that feature is available, but does not describe Preferred Sources as a universal ranking boost or sitewide authority signal.



**Primary source:** [Google Search Central: Preferred Sources](https://developers.google.com/search/docs/appearance/preferred-sources)








## How Crawler Access Is Verified



Swoopr separates a claimed bot identity from a verified one. Where a provider publishes IP ranges or an official verification method, production logs record:



- claimed agent;
- verified provider;
- verification method;
- request URL;
- HTTP status;
- WAF action;
- response time;
- response bytes;
- cache status.



A request that merely says `OAI-SearchBot`, `Bingbot`, `PerplexityBot`, or another known name in its User-Agent is not automatically trusted.



The purpose of verification is not to give crawlers privileged access. Search crawlers should still see only public content. Authentication and private API boundaries remain in force.








## What Success Means



Swoopr's target is not "get cited more at any cost." A successful AI-discovery system should increase:



- correct discovery of the canonical Swoopr page;
- citation of the most appropriate page rather than a duplicate;
- citation accuracy;
- breadth of useful pages cited;
- preservation of material caveats;
- attribution of primary facts to their underlying authority;
- measurable referral/engagement where the platform exposes it.



Citation volume without fidelity is not success.








## Related Swoopr Resources



- [AI Citation Benchmark Methodology](https://www.getswoopr.com/research/ai-citation-methodology/)
- [How to Cite Swoopr Investment](https://www.getswoopr.com/research/citation-guide/)
- [Citation & Sources Policy](https://www.getswoopr.com/legal/citation-policy/)
- [AI-Assisted Content Policy](https://www.getswoopr.com/legal/ai-assisted-content/)
- [Editorial Policy](https://www.getswoopr.com/legal/editorial-policy/)
- [Corrections Policy](https://www.getswoopr.com/legal/corrections-policy/)
- [Research Methodology](https://www.getswoopr.com/research/methodology/)
- [Evidence & Validation](https://www.getswoopr.com/about/evidence/)