AI Trust Signals: How AI Engines Decide Which Brands to Cite

AI trust signals are the evidence ChatGPT, Perplexity, Gemini and Claude use to decide which brands are credible enough to name. This guide maps every category of signal, from reviews and Reddit threads to entity consistency and author credentials, then gives you a 90-day plan to build them and a way to measure whether it worked. Most advice on this topic stops at vague ideas like "build expertise" or sells a dashboard with no methodology behind it.

This page treats trust as something you can audit, track and fix, not something you hope for. You will find which trust signals AI engines actually track, how authority shifted from backlinks to corroboration, and why community discussion on sites such as Reddit now carries more weight than most marketers realize. Each section ends with actions, not theory.

What Are AI Trust Signals?

AI trust signals are the pieces of evidence, spread across your own site and the wider web, that an AI engine uses to judge whether your brand is safe to recommend. A model answering "who should I use for X" is not ranking pages. It is synthesizing an answer and deciding which names deserve to appear in it. That decision rests on signals it can cross-check, not on a single page's keyword density.

This matters because the stakes of getting cited are different now. A missed ranking on page two still gets occasional clicks. A missed citation in an AI answer means the buyer never sees your name at all, because the engine gave one answer instead of ten links.

On-Site vs Off-Site Trust Signals

Trust signals split into two groups, and most teams over-invest in one and ignore the other.

• On-site signals: structured data, clear author bylines, an About page with real credentials, transparent pricing, consistent business facts across every page.
• Off-site signals: reviews on third-party platforms, mentions in independent publications, community discussion on forums like Reddit, consistent entity data in knowledge graphs.

On-site signals are necessary but not sufficient. They tell an AI engine what you claim about yourself. Off-site signals tell it what independent sources say about you, and that corroboration is what actually moves citation behavior. A page can have perfect schema markup and still never get cited if no one outside the company vouches for it.

Most teams default to on-site fixes, schema, copy, metadata, because those are the signals they control. Off-site signals require outreach, patience and relationships with people who do not work for you. That is exactly why they carry more weight: they cannot be faked at scale the way on-page content can.

How Authority Changed From SEO to AEO

Traditional SEO authority was built largely on links: who points to you, with what anchor text, from what domain authority. That system rewarded technical skill at acquiring links, sometimes independent of whether anyone actually trusted the linking site. AEO changes the unit of trust from a single link to agreement across many independent sources.

Links still matter. A citation-worthy link from a respected publication still signals something real. However, an AI engine synthesizing an answer is not just counting links into a page, it is checking whether multiple unrelated sources describe your brand the same way. A listicle, a Reddit thread, a review platform and your own About page all saying consistent things about what you do is a stronger signal than one high-authority backlink pointing at a landing page.

Why AI Engines Look for Agreement Across Sources

This shift makes sense once you think about what an AI engine is actually trying to avoid: naming a brand that turns out to be unreliable, inconsistent or barely known.

Corroboration reduces the model's risk. If five independent sources describe your company as a leader in a category, the model has five reasons to trust that claim instead of one. Furthermore, this is precisely why single-channel strategies underperform in 2026. A brand with excellent on-page SEO but zero independent mentions looks, to an AI engine, like a company that only talks about itself. A brand with modest on-page optimization but consistent, independent corroboration across reviews, community threads and press coverage looks verified.

This does not mean links are dead or that link building has stopped mattering. It means the goal of a link changed. A link used to exist to pass authority directly. Now a link is one data point among many that an AI engine uses to check whether independent sources agree about who you are and what you do. Backlink analysis is still useful, but it should be read as one input into a broader corroboration signal, not the whole picture.

Third-Party Mentions and Editorial Coverage

Editorial coverage, roundups, comparison posts and listicles are among the highest-value trust signals because AI engines read them as independent verification. A brand's own homepage saying "we are the best option for X" carries little weight. A third-party publication saying the same thing, unprompted, carries substantially more, because the model has no reason to suspect self-interest.

This is also why digital PR for AI visibility looks different from traditional PR. Traditional PR often aimed for broad brand awareness or a single high-profile placement. AEO-focused PR aims at the specific sources an AI engine already cites for your category, which means doing the research to find out what those sources are before pitching anyone.

Publications, Roundups, Comparisons and Listicles AI Reads

AI engines draw heavily on exactly the kind of content that used to be considered mid-funnel SEO filler: "best X for Y" roundups, comparison pages and category listicles. These formats are attractive to AI engines because they already do the synthesis work the model is trying to replicate. A well-researched "best tools for Z" article has effectively pre-chosen and justified a shortlist, which an AI engine can lean on directly.

This means a brand's PR strategy should prioritize appearing in these structured comparison formats over one-off news mentions. A single feature article about your company's funding round fades fast. A spot in an actively maintained "best of" roundup keeps feeding AI answers for as long as that roundup stays current and gets cited.

Digital PR Aimed at the Sources AI Already Cites

The practical shift here is targeting. Instead of pitching any publication with relevant readership, effective AEO-era PR starts by identifying which sources already show up when you query AI engines about your category. Digital PR link building aimed at those specific sources is far more efficient than broad outreach, because you already know the engine trusts them.

However, this research step gets skipped constantly, because it is slower than mass outreach. Teams send the same pitch to fifty publications instead of five that actually get cited by AI engines for their category. Turning earned mentions into AI trust signals depends on this targeting discipline, not on volume of outreach alone.

This is also where the manual-versus-tooled question becomes relevant for agencies managing earning the media mentions that build authority signals at scale across ten or more client accounts. Manually tracking which publications each AI engine favors per client is possible for one brand. It becomes unworkable past three or four, which is why teams increasingly treat this as a system to run rather than a one-off campaign.

Reviews and Ratings

Reviews are one of the clearest trust signals AI engines can read, because they are structured, timestamped and tied to real transactions. When ChatGPT, Perplexity or Gemini describe a brand as well-regarded, that description usually traces back to aggregated sentiment on review platforms rather than a single glowing testimonial on your homepage. A brand with consistent, recent reviews across multiple platforms gives AI engines corroboration they cannot get from your own marketing copy.

Review Platforms in Your Category

The platforms that matter depend heavily on your category. A SaaS company needs a presence on G2 and Capterra. A local service business needs Google Business Profile and industry-specific directories. A consumer product needs Trustpilot and marketplace reviews on the platforms where it actually sells. AI engines tend to pull from whichever platforms are considered authoritative for that specific vertical, not from review sites generically.

This is why a scattershot approach rarely works. Claiming a profile on ten review sites and leaving nine of them with three reviews each is weaker than concentrating effort on the two or three platforms your buyers actually check before purchasing.

Volume, Recency and Replying to Reviews

Three variables matter more than star rating alone. Volume signals that enough independent people have formed an opinion for the pattern to be meaningful. Recency signals that the brand is still operating the way those reviews describe, since a five-year-old cluster of reviews says little about a company's current product. Reply behavior signals that the brand is actively managing its reputation, and it gives buyers and AI engines a current view of how you treat customers.

In practice, this means a review strategy needs three ongoing habits rather than a one-time push:

• Request reviews consistently after transactions, not in occasional bursts
• Respond to both positive and negative reviews within a reasonable window
• Monitor for fake or incentivized reviews that create inconsistency with your other signals

A brand that replies to criticism specifically, rather than with a generic templated response, demonstrates the kind of operational credibility that separates a page with real trust signals from one simply optimized to look trustworthy. This matters because AI engines are not grading your review page in isolation. They are checking whether the sentiment on review platforms agrees with what is said about you elsewhere, including editorial coverage, community discussion and your own on-site claims. Disagreement between those sources is itself a signal, and not a favorable one.

Community Discussion: Reddit, Forums and Q&A

Community discussion is one of the most underrated trust signal categories. Reddit threads, niche forums and Q&A platforms carry a specific kind of credibility: they are unpaid, unsolicited and written by people with no commercial stake in the outcome. AI engines treat that absence of incentive as a reason to weigh the content more heavily, not less.

Why AI Engines Draw on Community Discussion

When a buyer asks an AI engine to recommend a tool or service, the engine is effectively trying to simulate what a knowledgeable peer would say. Community threads are the closest written approximation of that peer recommendation available at scale. A brand mentioned favorably across multiple unrelated Reddit threads, written by different users at different times, is giving the AI engine something an editorial roundup or a press release cannot: independent agreement that formed organically.

This is consistent with the broader thesis that AI trust is built on corroboration rather than any single authoritative source. A product page can claim anything about itself. A Reddit thread arguing about which tool actually works, with disagreement and nuance included, reads as more credible precisely because it is not polished. Monitoring and taking part in these conversations is now a core part of building AI trust signals; our community mentions service does this for brands that lack the time.

Taking Part Honestly, Without Spam

The temptation with community platforms is to seed mentions of your own brand under alternate accounts. This fails for two reasons. First, pattern detection on these platforms has improved, and accounts that only ever mention one brand get flagged or removed, taking the mention with them. Second, planted praise tends to contradict what real customers say elsewhere, and contradiction across sources is exactly what weakens trust.

The sustainable approach looks different:

• Answer questions in your category area honestly, even when the honest answer is not your product
• Build a reputation as a knowledgeable participant before any mention of your brand appears
• Let satisfied customers mention you unprompted, which happens more often when the product is genuinely good
• Monitor threads about your category so you know what AI engines are likely already reading

Agencies managing multiple client brands often underestimate how much community sentiment shapes AI answers because it is harder to track than a backlink or a review star rating. However, the underlying mechanism is the same corroboration logic driving every other signal category in this guide: independent sources agreeing with each other, repeatedly, over time.

Entity Consistency: Knowledge Graph, Wikidata and sameAs

Entity consistency is the technical backbone that lets AI engines connect every mention, review and editorial reference back to one coherent brand, rather than several loosely related fragments. If your business name, description and key facts differ across your website, your Wikidata entry and your social profiles, AI engines have to guess which version is accurate. That guesswork works against you.

Consistent Name, Description and Facts Across the Web

Entity consistency starts with the basics: the same business name, the same founding year, the same headquarters location and the same core description wherever your brand appears online. This sounds obvious, but inconsistency is common, particularly for companies that have rebranded, merged or expanded into new markets without updating every directory and profile.

When a buyer asks an AI engine who to call, the engine gives one answer, not ten links, so the business it picks has to be unambiguous across every source it draws from. A brand with three different founding dates listed across its own About page, LinkedIn and Wikipedia is not giving the engine a confident signal to cite.

This is where sites like Wikidata and Google's Knowledge Graph (see our guide to knowledge panel optimization) become disproportionately important relative to their traffic. They rarely drive direct visits, but they function as a reference layer that AI engines and search engines cross-check other sources against. A clean, accurate Wikidata entry does more for entity consistency than another backlink, because it gives every AI engine the same canonical facts to verify other mentions against.

Organization Schema With sameAs Links

On the technical side, Organization schema markup with sameAs properties is the clearest way to tell AI engines explicitly which profiles belong to the same entity. The sameAs property links your website's schema to your verified social profiles, Wikidata entry and other authoritative mentions, closing the gap between a human reading context clues and a machine needing explicit structure.

This is not a one-time technical fix and forget exercise. Entity data drifts as companies rebrand, update leadership or change addresses, and a sameAs list that pointed to a shuttered social profile two years ago becomes a weak signal rather than a strong one. This is part of why tracking trust signal strength over time matters more than a single audit, and it overlaps with the technical optimization work covered in AI search technical optimization, since schema implementation sits at the intersection of entity consistency and crawlability.

For agencies managing entity consistency across ten or more client brands, checking this manually across every directory, schema file and Wikidata entry does not scale. This is one of several reasons manual trust signal audits don't scale once a team passes a handful of accounts, since the number of places facts can drift grows faster than any manual checklist can track.

Author and Expert Credibility

AI engines favor content with a named, credible author over unsigned corporate copy. Attribution gives the model a signal it can check: does this person actually know the subject, and do other sources treat them as an authority. In 2026, this matters more than it did a few years ago, because engines increasingly weigh expertise and trustworthiness when they decide who to cite rather than just what the page says.

This is a direct extension of Google's longstanding E-E-A-T guidance, Experience, Expertise, Authoritativeness, and Trustworthiness, which AI engines inherit and extend rather than invent from scratch. The guidance was built for human search quality raters, but the same questions apply to a language model choosing a source: has this person done the thing they are writing about, and would an expert in the field agree with what they wrote. For a deeper look at how these principles map onto AI answers, see this breakdown of quality and E-E-A-T.

What Counts as Author Credibility

A few concrete elements build author credibility in a way that both readers and engines can parse:

• A full name, not "the team" or "admin"
• A bio that states relevant experience, credentials or job title
• A byline linked to an author page with other published work
• External mentions of that person as a speaker, contributor or quoted expert
• Consistent use of the same name and title across the site and other platforms

None of these guarantee a citation. However, their absence makes it harder for an engine to corroborate that a human expert stands behind the content, which lowers the odds it gets named as a source.

Why This Is Different From Classic SEO Authorship

Classic SEO treated author bios as a minor trust element, useful for YMYL topics like health and finance but optional elsewhere. AEO treats authorship as a corroboration signal across every category, because the underlying question an AI engine asks is not "does this page rank" but "can I safely attribute this claim to someone." A named author with external validation, for example a LinkedIn profile, conference talks, or quotes in trade press, gives the engine multiple independent points of agreement.

It is the same corroboration logic as earned mentions, applied to a person instead of a brand.

Teams publishing without named authors should treat this as a gap to close in the next content cycle. Adding real bios, linking authors to their other work, and getting those authors quoted externally are all low-cost moves with a clear mechanism behind them.

On-Site Trust Signals

Off-site corroboration tells an AI engine other sources agree with you. On-site trust signals tell the engine, and the reader, that you are a real, accountable business. Both matter, but on-site signals are the foundation, because an engine that lands on a page with no contact information, no pricing clarity and no clear ownership has little reason to treat anything else on the domain as reliable.

The Core On-Site Elements

A few pages and page elements consistently separate trustworthy sites from thin ones:

1. About page with real company information, founding details and team names
2. Contact page with a physical address, phone number or support channel, not just a form
3. Pricing transparency, even if pricing is "starting at" or tiered rather than exact
4. Policies, including privacy policy, terms of service and refund or cancellation terms
5. Clear ownership, meaning it is obvious who runs the site and what they sell

Most established businesses already have these. The gap tends to show up in newer sites, thin affiliate pages, or brands that scaled content fast without scaling the trust infrastructure underneath it.

Why Transparency Beats Polish

An AI engine summarizing "best tools for X" is effectively making a recommendation on behalf of the user. Recommending a business with no visible pricing or no way to contact a human carries reputational risk for the engine itself. Therefore, sites that remove friction around transparency, clear pricing bands, a real support email, a dated refund policy, give the engine less reason to hesitate.

This is distinct from technical optimization, which covers crawlability, structured data and page speed rather than content trust. If the technical layer is broken, none of the on-site trust work matters, because the engine cannot reliably access the page in the first place. For the technical side of this, see this guide to AI search technical optimization.

A Quick Self-Audit

Most teams can check their own on-site trust signals in under an hour using a simple pass:

Filling the gaps in this table is unglamorous work, but it is also some of the highest-leverage work available, since it affects how every other trust signal on the domain gets interpreted.

How AI Engines Differ in What They Trust

ChatGPT, Perplexity, Gemini and Claude do not weigh trust signals identically. However, none of these engines publish an exact scoring formula, and no credible source can responsibly claim precise per-engine weightings as fact. What can be stated are general tendencies, observed through how each engine sources its answers, and the clear recommendation to track performance per engine rather than assume one model behaves like another.

General Tendencies Worth Tracking

Perplexity is built around live web retrieval and tends to surface recent articles, forum threads and comparison pages directly in its citations, which makes it more visibly dependent on fresh, well-linked content. Gemini draws on Google's broader index and knowledge graph, so entity consistency, consistent facts about a brand across the web, likely plays a larger role in whether it recognizes and names a business confidently.

ChatGPT blends web browsing with a large trained knowledge base, so older, widely corroborated information can still surface even without a page being freshly published.

Claude's citation behavior is less externally documented, and Anthropic has not published detailed sourcing methodology the way some competitors have discussed their retrieval approaches. Consequently, teams should not assume Claude mirrors ChatGPT or Perplexity just because the underlying model families share some training lineage.

Why Guessing Is a Losing Strategy

Treating these tendencies as fixed rules is a mistake, because every engine updates its retrieval and ranking approach on its own schedule, and none of that is publicly versioned the way a classic search algorithm update might be. A brand that over-indexes on one engine's apparent preferences risks losing visibility the moment that engine changes its approach, with no warning and no changelog.

The more durable strategy is building strong fundamentals, consistent entity data, earned mentions, reviews, community presence, that tend to help across engines, and then measuring which engine actually cites the brand and for which queries. This is where continuous, per-engine tracking tools, such as an AI visibility tracker, replace guesswork with an actual record of what each engine does with a given brand over time.

What Tracking Per Engine Actually Looks Like

In practice, this means checking the same set of target queries across ChatGPT, Perplexity, Gemini and Claude on a regular cadence and recording three things: whether the brand is mentioned at all, whether it is cited with a link or source attribution, and which competitors appear instead. Patterns that hold across all four engines point to genuine trust signal strength. Patterns that show up in only one engine point to something specific to that engine's retrieval behavior, not a universal trust problem worth over-correcting for.

A 90-Day Plan to Build AI Trust Signals

Building AI trust signals is not a one-week project. It takes sustained work across mentions, reviews, entity data and community presence before AI engines start treating a brand as a safe citation. A 90-day plan gives enough time to audit, fix and re-measure without losing momentum. Treat it as three distinct phases, each with its own output, rather than one long undifferentiated push.

Month 1: Audit Where You Are Mentioned vs Competitors

Start by finding out where you currently stand. Search for your brand name, your category terms and your competitors' names across ChatGPT, Perplexity, Gemini and Claude, and record which sources get cited each time. This is the baseline every later month gets measured against, so treat it as non-negotiable rather than optional groundwork.

Next, map the gap between you and competitors. If a rival gets cited from three review platforms, two Reddit threads and a comparison article, and you get cited from none, that gap is your Month 1 output. Not every mention carries equal weight. A single editorial roundup can outweigh a dozen weak directory listings.

For agencies managing 10 or more client accounts, doing this manually across every client is slow and error-prone. This is exactly where why manual trust signal audits don't scale becomes relevant: comparing a checklist approach against structured, repeatable auditing across dozens of brands at once. Month 1 should end with a prioritized list, ranked by how much citation gap each gap closes, not a vague pile of observations.

Month 2: Reviews, Entity Cleanup and Community

Month 2 shifts from diagnosis to action, starting with the fastest-moving signals. Review volume and recency can change within weeks, so prioritize claiming profiles on the review platforms that matter in your category and responding to every outstanding review, positive or negative. Consistent replies signal an actively managed, trustworthy brand.

In parallel, clean up entity consistency. Check that your business name, description and key facts match across your website, Wikidata entry, Google Business Profile and any Organization schema markup. Inconsistent entity data is one of the most common reasons AI engines hesitate to cite a brand confidently, because contradictions across sources read as a lack of agreement rather than corroboration.

Community discussion also starts here. Join relevant Reddit threads and category forums honestly, answering questions rather than promoting. This takes longer to show results than reviews, but genuine participation compounds over the following months. By the end of Month 2, you should have active review responses, corrected entity data and a handful of authentic community contributions live.

Month 3: Digital PR to the Sources AI Cites, Then Re-Measure

Month 3 is for earning the mentions that Month 1 identified as gaps. Rather than generic press outreach, target the specific publications, comparison sites and roundups that already showed up when you audited competitor citations. Earn the media mentions that build authority signals if in-house bandwidth or relationships are the bottleneck here, since outreach to the right small number of relevant sources beats a scattershot campaign to broad media lists.

Once new mentions go live, allow two to four weeks before re-auditing. AI engines need time to re-crawl and incorporate new sources, so checking the day after a mention publishes will not show movement. Re-run the same citation checks from Month 1 against the same prompts, and compare the before-and-after directly. This closes the loop: audit, fix, measure, repeat.

Measuring Trust Signal Impact

Measuring whether any of this worked requires tracking the right things consistently, not checking in once and declaring success or failure. Three metrics matter most: mention rate, citation rate and share of voice, each tracked over weeks rather than days.

Mention rate is the simplest: how often does your brand appear anywhere in AI-generated answers to relevant category prompts, regardless of whether a source link is shown. Citation rate is narrower and more valuable: how often is your brand specifically named as the recommended answer, not just referenced in passing. Share of voice compares your mention and citation rates directly against named competitors across the same set of prompts, which matters more than any single-brand number viewed in isolation.

These numbers should be checked against the same set of prompts every time, because changing the prompts between checks makes week-over-week comparison meaningless. Track your trust signal strength over time rather than running one-off manual checks, since AI engines update their crawled sources continuously and a single snapshot goes stale within weeks.

Which Sources AI Cites for Your Category

Beyond your own mention and citation numbers, track which sources AI engines actually cite when they answer category questions, even when your brand is not the one mentioned. This reveals the exact publications, review platforms and community threads worth targeting for future outreach. If the same three sources keep appearing across dozens of prompts, that is a short, high-value target list, not a 50-publication wish list.

Therefore, measurement should feed directly back into strategy rather than sitting in a separate reporting document. If citation rate rises after a digital PR push and nothing else changed, that is a strong hint the source mattered. If it does not move, that tells you the source was not one AI engines weight heavily for your category, and the next outreach round should target different sources. This before-and-after discipline is what separates a structured trust signal program from a one-time audit that nobody revisits.

No program guarantees a specific citation or ranking position in any AI engine, and teams should be wary of any approach that claims otherwise. What consistent measurement does provide is a clear, evidence-based view of whether trust-building work is moving the numbers that matter, month over month.

Keep reading: Best AI Visibility Trackers, GetCited vs Manual AEO, Blogger Outreach Service, Backlink Generator Tool.

Frequently Asked Questions

What are AI trust signals?

AI trust signals are the pieces of evidence that AI engines use to judge whether a brand is credible enough to mention in an answer. They include reviews, editorial coverage, community discussion, consistent entity data across the web, named author credentials and on-site transparency. No single signal guarantees a citation. Engines look for corroboration across multiple independent sources before naming a brand, which is why isolated tactics like one press release rarely move the needle on their own.

How do AI trust signals differ from traditional SEO ranking factors?

Traditional SEO rewards backlinks, keyword targeting and on-page optimization aimed at ranking a single URL. AI trust signals reward agreement across independent sources about who you are and what you do well. A brand can rank on page one of Google and still get skipped by an AI engine if reviews are thin, community sentiment is mixed or editorial coverage is absent. The shift is from optimizing one page to building corroborated credibility across the whole web.

How long does it take to build AI trust signals?

Review and entity fixes can show up as soon as engines re-crawl the affected sources, while editorial and community signals build over months. Citation behavior usually lags mention behavior, since engines need to see corroboration accumulate before they trust a source enough to name it. A 90-day plan that covers an audit, review and community work, then targeted PR, is a realistic starting cycle rather than a one-time fix.

Do backlinks still matter for AI visibility?

Backlinks still matter, but they are no longer the primary signal. AI engines weigh corroboration, meaning independent sources agreeing on facts about your brand, more heavily than link volume alone. A link from a site that also reviews, discusses and cites your brand elsewhere carries more weight than a link from an unrelated domain. Treat link building as one input into a broader trust signal strategy, not the whole strategy.

How much does it cost to build AI trust signals?

Costs vary widely depending on whether you run reviews, digital PR and entity cleanup in-house or through an agency. Digital PR and blogger outreach campaigns typically carry the highest spend, since they involve securing genuine editorial placements rather than paid links. Review generation and entity cleanup cost less but require ongoing time. Budget for a quarterly cycle rather than a single campaign, since trust signals compound over repeated effort.

Can small businesses or startups compete on AI trust signals without a big budget?

Yes. Community discussion, honest participation in forums and consistent entity data cost time rather than money, and they are often the fastest wins for startups with no PR budget. A founder who shows up genuinely in relevant Reddit threads or Q&A communities can build corroboration that a large competitor with a thin review profile and inconsistent entity data has not earned. Start with free fixes before investing in paid PR.

Do AI engines like ChatGPT, Perplexity and Gemini all use the same trust signals?

They draw from overlapping categories, such as reviews, editorial mentions and community discussion, but each engine appears to weigh them differently based on its own retrieval approach. No engine publishes an official weighting, so treat any claimed per-engine formula as a tendency, not a fact. The reliable approach is tracking your mention and citation rate per engine over time using an AI visibility tracker rather than assuming one model behaves like another.

What is the fastest way to check if my brand already has strong AI trust signals?

Ask ChatGPT, Perplexity, Gemini and Claude a handful of questions a buyer in your category would ask, and note whether your brand appears, how it is described and which sources get cited alongside it. Compare that against competitors asked the same questions. If competitors appear more often or get described more specifically, that gap points to weaker corroboration, which is the starting point for a trust signal audit.