AI search technology is advancing quickly, not just in terms of capabilities, but in its processes as well. The signals AI relies on to decide who to cite are constantly shifting. This means optimisation approaches are also changing, with new ideas regularly surfacing online. 

At TDMP, we notice plenty of hype around the latest “silver bullets” for boosting AI visibility, but caution is important. Instead of jumping on every emerging trend, we take a pragmatic approach, researching and testing developments thoroughly before implementing what we think is best for our clients.

Here is a look at what has changed in the field of AI visibility recently, why AI answers behave the way they do, and how we are adapting our strategies to futureproof client success.

A quick primer on the impact of AI search

• AI answer engines (ChatGPT, Gemini, Claude, etc.) give direct, conversational answers to searches, which means users no longer have to visit websites to find answers and carry out research.
• Clicks from search engine results pages (SERPs) are down across the board, and the zero-click trend is accelerating.
• Brands now need to build their presence in AI-generated answers to continue reaching target audiences at scale.
• Effective SEO is pivoting from Search Engine Optimisation to Search Everywhere Optimisation, supporting visibility wherever search happens - be it Google’s traditional organic ranks, AI Overviews, or standalone AI answer engines.

What’s changed in AI visibility?

Schema is no longer optional for your AI visibility strategies

For years, Schema markup and structured data were considered SEO nice-to-haves, useful for earning rich snippets, but rarely a dealbreaker for securing high search engine rankings.

In AI-driven search, this has changed. Schema is now a hard requirement for optimising for AI visibility.

AI models don't read web pages the way humans do; they require clear page hierarchy and machine-readable data to understand context, entities, and relationships. This doesn’t mean you need to prepare dedicated markdown versions of your pages for AI consumption, but it does mean structured data is essential. 

If your structured data is incorrect, missing, or broken, aspects of your business may be invisible to AI engines.

Even if you follow SEO best practices by the book and your webpages rank well on traditional blue-link search results pages, the absence of appropriate schema can limit your visibility in AI search interactions.

The standards landscape is fragmented (and that’s okay)

Whenever search changes, a wave of new protocols and formats compete for adoption.

For instance, Google recently introduced proposals for the Open Knowledge Format (OKF), an open initiative designed to make structured data easier for various AI agents (including those from OpenAI, Anthropic, and Microsoft) to read directly. At the same time, platforms like Cloudflare are proposing their own approaches to AI web crawling.

Related - Robots.txt, AI crawlers - and what (if anything) you should be doing about it

Not all of these standards will survive, and we don't believe in chasing every new format that hits the headlines.

What initiatives like OKF do prove is that our pragmatic, measured approach is the right one. AI systems are moving away from simple web page retrieval toward machine-readable knowledge networks. Whether a specific format wins out or not, the underlying requirement remains the same: clean structure, verified facts, and clear relationships between your data.

Addressing AI idiosyncrasies causing confusion for brands

Why are AI-generated answers always changing?

A question we frequently get from clients is: "Why does AI say something different about our business every time we ask?" or “Why is that the AI recommended us the first time we used this prompt, but then didn’t mention us when we used the same prompt again?”.

It isn't a bug, and it doesn’t mean that you were doing something right and then got it wrong somewhere along the way. This is just how the technology works.

The reasons an AI answer will differ every time, even for identical prompts, are as follows:

1. Inherent technical variability: To handle millions of questions at once, AI platforms group your prompt together with requests from other users and processes them simultaneously. The “batch-size” changes with every prompt based on use demand at the moment of search, and this shifting workload causes the computer to calculate its math in slightly different orders each time, introducing subtle variations.
2. The snowball effect: LLMs generate responses word-by-word based on probabilities. If a tiny math variation due to differing batch sizes flips just one word at the start of a sentence, the entire trajectory of the rest of the answer may change.
3. Personalisation & context: AI search engines tailor responses based on user location, search history, and previous chat context.
4. Continuous updates: Models, web indexes, and retrieval algorithms are updated constantly behind the scenes, shifting how data is weighted.

No two answers will look identical. Because of this, trying to rank for an exact, fixed phrase in an AI answer is the wrong goal. The goal is ensuring the AI has access to accurate, consistent facts about your business whenever it generates a response.

This increases the likelihood that your business, offerings or website is incorporated into the answer across a range of relevant prompts, but it does not provide the same level of consistency you may be accustomed to in traditional SEO.

Why does AI say unexpected things about brands?

If an AI platform says something unexpected about your brand, it’s likely because it isn't looking solely at your website. 

AI search platforms aim to surface the most trusted, most-cited sources for a given topic. Sometimes that source is your website. Often, it’s a third-party reference, an industry directory, or a database in which the AI has confidence.

“Confidence”, in this context, means the AI has come across very few information conflicts about the source in question and the entity is validated by authoritative third-party sources across the web. 

This is why entity clarity is central to supporting visibility in AI answers. Entity clarity means AI has a crystal-clear understanding of who you are in contrast to everything else on the web.

For example, our sister company, Adaptive, need to ensure the AI explicitly distinguishes them as their specific business, rather than:

• The general concept of "adaptiveness" (the dictionary definition).
• The thousands of other businesses that happen to use "Adaptive" in their name.

This is achieved by sending hyper-consistent signals across a brand’s entire digital footprint. Without strong entity clarity, the AI gets confused about which business to pull data for. The more conflicts it finds, the less confidence it has, and the lower the chances it will cite or recommend your business.

What do AI visibility strategies look like in SEO moving forward

Effective AI visibility strategies are a natural evolution of the SEO principles we’ve advocated for years, such as E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness), topic clusters, and technical precision.

The focus is just expanding:

• From keywords to entities: AI needs to understand what your business, products, and services are, not just match search queries to words on a page.
• From isolated pages to connected information: Your case studies, FAQs, author bios, and NAP (Name, Address, Phone) data all contribute to a single picture of authority and contribute towards entity clarity.
• From ranking a page to managing a Knowledge Graph: Ensuring your brand's data is complete, consistent, and trusted enough for AI engines to reference confidently.

As for reporting, search visibility analytics are still developing, with both Google and Bing having recently launched AI visibility reports in Search Console and Bing Webmaster Tools respectively.

The TDMP approach: what we’re doing for our clients

Rather than completely reinventing the wheel, we are integrating AI readiness directly into our day-to-day work:

• Schema reviews & scoring: We’ve built structured data checks and technical schema scoring into our regular reporting cycles to ensure no client is left invisible.
• Entity & Wikidata management: We proactively manage connected data signals, including Wikidata; Google Business Profiles; and structured NAP data, to support Google’s Knowledge Graph and feed LLMs accurate information.
• AI Readiness framework: To identify clear gaps and prioritise fixes that deliver practical value, we assess client sites against three core pillars:
  • Trust
  • Relevance
  • Usability

Is your business visible to AI?

We're currently rolling out AI Readiness reviews across our active accounts. If you'd like to see how your site is currently interpreted by AI engines and where the immediate opportunities are, let’s talk.