Popular on Ex Nihilo Magazine

Marketing & Growth

AI Doesn’t Care About Your Testimonials

You spent hours collecting testimonials. Putting them on a dedicated page with photos and job titles. You rotated the

AI Doesn’t Care About Your Testimonials

You spent hours collecting testimonials. Putting them on a dedicated page with photos and job titles. You rotated the best ones onto your homepage. Every marketing guide told you this builds trust.

AI doesn’t care.

When ChatGPT, Perplexity, or Claude decide which businesses to recommend, your testimonials page carries almost no weight. The same words, from the same client, posted on a platform like Trustpilot or Google Reviews, count for significantly more. The content is identical. The credibility is not.

This is one of several findings from a March 2026 research paper by the Scientific Institute for Generative Intelligence (SIGI), which analysed how large language models evaluate business credibility. The researchers found that verified external signals consistently outperform self-reported claims, even when the information is the same. For anyone thinking about ChatGPT SEO or generative engine optimisation (GEO), this distinction matters.

Trustpilot Reviews Score 8.0. Your Testimonials Score 5.5.

The SIGI research scored 77 different trust signals that influence whether AI recommends a business. Platform-verified reviews scored 8.0 out of 10. Self-hosted testimonials scored 5.5.

The difference comes down to verification. Your testimonials page is something you control completely. You choose which reviews to display. You can write them yourself. There’s no verification step.

Platform-verified reviews work differently. Trustpilot interviews reviewers. Google connects reviews to actual transactions. The platform sits between you and the testimonial, adding a layer of independent verification.

The SIGI model reported that this verification step adds credibility that self-reported testimonials cannot match. Even when the testimonial content is identical.

For ChatGPT SEO, this creates an uncomfortable truth: the testimonials you’ve collected may help human visitors who land on your site, but they do almost nothing to help AI systems find you in the first place.

Businesses on 4+ Platforms Are 2.8x More Likely to Appear

The highest-scoring social proof signal wasn’t reviews at all. It was multi-platform presence.

According to the SIGI research, businesses present on four or more platforms are 2.8 times more likely to appear in AI responses. The researchers described this as a consensus signal. When multiple independent sources confirm that a business exists and does what it claims, AI systems treat that as corroboration.

Consider what ChatGPT is doing when someone asks “who are the best accountants in Manchester.” It can’t visit offices, check credentials, or ask around. It relies on signals in its training data and retrieval systems. A business that appears on Google Business Profile, LinkedIn, Trustpilot, and an industry directory creates four independent data points. A business with only a website creates one.

Anyone building a ChatGPT SEO strategy should start here. External validation beats self-assertion. Consensus beats isolated claims.

Five Named Clients Beat “80+ Clients”

Named client portfolios scored 7.0 in the SIGI research. Anonymous client counts scored substantially lower.

The difference is verifiability. When a web design agency says “we’ve worked with Deliveroo, Monzo, and Oatly,” that’s checkable. Someone could verify those relationships. When the same agency says “trusted by 80+ clients,” that’s a claim without evidence. It might be true. But AI systems have no way to confirm it, so they discount it.

The research found that named clients are “dramatically more credible” than anonymous counts. Listing five real clients beats claiming eighty anonymous ones.

This inverts how many businesses approach ChatGPT SEO. The instinct is to sound impressive through volume. But volume without names is noise to an AI system trying to verify what’s real.

Wikidata Takes 30 Minutes and Almost Nobody Does It

One of the stranger findings involves Wikidata, the structured database that feeds Google’s Knowledge Graph. Wikidata contains information on approximately five billion entities and serves as a primary source for AI systems evaluating whether something is a recognised entity.

The SIGI research found that Wikidata presence confers “recognised entity status in the global knowledge base.” Being listed there signals to AI systems that you’re a real entity worth knowing about.

Almost no small businesses are in Wikidata. Creating an entry takes about 30 minutes and the effect is permanent. The research described this as a first-mover advantage for businesses willing to do something their competitors haven’t thought of yet.

Wikidata has notability requirements, but they’re more flexible than Wikipedia’s. A business with meaningful coverage in reliable sources can often qualify. For generative engine optimisation, this is one of the highest-impact actions relative to effort.

ChatGPT SEO Is About Verification, Not Volume

Traditional SEO rewarded volume: more pages, more backlinks, more keywords. ChatGPT SEO and GEO reward verification: more independent sources confirming the same facts about your business.

Platform-verified reviews beat self-hosted testimonials. Multi-platform presence beats a single website. Named clients beat anonymous counts. Wikidata entries beat obscurity.

None of this replaces Google. Traditional search still matters and will for years. But ChatGPT has over 100 million weekly users. Perplexity is growing fast. When someone asks an AI for recommendations, your testimonials page won’t help you appear in the answer.

The fix is straightforward. Get verified reviews on platforms that verify. Maintain active profiles across multiple directories. Name your clients instead of counting them. Spend 30 minutes creating a Wikidata entry.

Your testimonials page can stay. It might still help humans who find you through other means. But for ChatGPT SEO, AI has already decided it’s not credible evidence.

Sources:

SIGI: Social Proof Signals in AI Evaluation: From Verified Reviews to Multi-Platform Presence

arXiv: GEO: Generative Engine Optimization


Ex Nihilo magazine is for entrepreneurs and startups, connecting them with investors and fueling the global entrepreneur movement

About Author

Conor Healy

Conor Timothy Healy is a Brand Specialist at Tokyo Design Studio Australia and contributor to Ex Nihilo Magazine and Design Magazine.

Leave a Reply

Your email address will not be published. Required fields are marked *