How to Optimise Content for AI Search: What the Technical Guides Miss (and What Actually Gets You Cited)

Optimising content for AI search — also called generative engine optimisation (GEO) — means building content that AI systems like Google AI Overviews, ChatGPT, and Perplexity cite when generating answers for users. The short version of what it requires: content that demonstrates genuine expertise, is clearly structured, uses schema markup correctly, and answers specific questions directly. Done well, it means your firm appears not just on a results page but in the synthesised answer a potential client receives before they’ve clicked anything.

For professional services firms — solicitors, IFAs, mortgage brokers, chartered surveyors, management consultants — this matters more than most marketing channels right now. But most of the guidance written about it misses the real challenge. The technical fixes are easy to implement. What they can’t fix is a thin content library that AI models have no reason to cite.

This guide covers both layers: what makes content worth citing in the first place, and the technical signals that make citable content extractable.

---

Genuine content depth (primary)	AI systems cite sources that demonstrate expertise — thin content earns no citations regardless of technical signals
Structured formatting (H2/H3, bullet points, FAQ blocks)	Makes content extractable — AI models parse structure, not prose
Schema markup (Article, FAQPage, BreadcrumbList)	Signals content type and purpose to AI crawlers
Inline expert citations and sourced claims	Builds the authority signals AI systems weight heavily
Content accuracy, not freshness	AI Overviews are not recency-biased — the median cited page is 14 months old (Digital Applied, April 2026); maintaining accurate, substantive content matters more than publishing frequency

A note on who produced this: SwyftSystems builds AI-search-optimised content for specialist professional services firms. The methodology in this guide is the same methodology we use in production — which means this article is a worked example of the approach we’re describing. You can evaluate both simultaneously. We work exclusively in professional services, and we’ll say directly if a firm’s situation isn’t one our system is designed for.

Why AI search is now relevant to professional services firms

What changed at Google I/O 2026

In May 2026, Google declared AI Mode the most significant change to Search in 25 years. The transition isn’t coming — it’s already happening. According to a 2025 Semrush study of Google AI Overviews, the feature now appears in approximately 88% of informational search queries. For question-format searches — precisely the queries a potential client runs when evaluating whether to hire a solicitor, engage an IFA, or instruct a surveyor — AI search has become the default delivery mechanism for answers.

That means when a business owner searches “how do law firms get clients through SEO” or “should I hire a content agency or build an in-house team,” the answer they receive first is generated by an AI, synthesised from sources the model has decided are credible. If your firm’s content isn’t one of those sources, you’re not in the conversation.

What “being cited” actually means — and how it differs from ranking

Traditional SEO puts your article at position 3 on a results page. AI search citation means your content is one of four or five sources an AI model draws on when writing a generated answer — and your site is listed as a reference. The reader sees the synthesised response first. They then decide whether to click through for more.

Research from Ahrefs (March 2026, 863K SERPs) found that 37.9% of URLs cited in AI Overviews rank in the top ten organically — down sharply from 76% in Ahrefs’ July 2025 data. Ahrefs attributes the drop partly to Google’s query fan-out process, where an initial query is expanded into multiple sub-queries and citations are drawn from across those results, not just the original SERP. Strong organic performance still contributes, but citation is increasingly decoupled from ranking position. That’s what creates the opportunity for specialist firms that have built genuinely substantive content: they can be cited over larger, more established sites if their content better demonstrates expertise on the specific question being asked.

What AI systems actually look for when deciding what to cite

E-E-A-T is the entry requirement, not the differentiator

Google’s framework for evaluating content quality — Experience, Expertise, Authoritativeness, and Trustworthiness (E-E-A-T) — applies to AI Overviews as Google Search Central documentation confirms (developers.google.com, verified May 2026). AI models from other platforms apply similar signals under different names. The practical implication: content that demonstrates no real expertise won’t be cited, regardless of how well it’s structured technically.

For professional services firms, this is simultaneously an advantage and a challenge. The expertise is genuine — your solicitors know property law; your IFAs understand pension transfer analysis; your surveyors can explain defect classification. The problem is that expertise locked in people’s heads isn’t visible to AI models. It has to be published, in substantive articles that demonstrate the thinking, not just assert the conclusion.

Content depth: why thin articles won’t be cited regardless of technical signals

This is the gap most guides don’t address. You can add perfect schema markup to a 600-word overview of a topic and it still won’t earn citations. AI systems evaluate whether the content genuinely answers the question a user asked — and whether it answers it better than the alternatives. A thin article, no matter how well formatted, loses that comparison.

What “depth” means in practice:

• The article fully answers the question the headline promises to answer, not a watered-down version of it
• It addresses the specific context your reader is in — a solicitor asking about content marketing isn’t the same as a general business owner asking the same question
• It includes specific claims, not general statements (“firms that publish consistently tend to outperform firms that publish sporadically” is stronger than “publishing content is important”)
• It cites sources for factual claims — research findings, regulatory references, industry data

The depth layer is what most professional services firms haven’t built yet. It’s also the layer that does the most work.

Structure and schema: the signals that make citable content extractable

Once you’ve built content with genuine depth, the technical signals determine how well AI systems can extract and use it. According to Digital Applied’s 1,000 AIO Citation Study (April 2026), schema-marked pages are cited 2.3× more often than otherwise-comparable unstructured pages after controlling for domain authority. Pages with clear H2/H3 structures are significantly more likely to be cited than pages with the same information buried in undifferentiated prose.

This isn’t because AI models can’t read prose — it’s because structure makes the relevance of each section immediately identifiable. An H2 that reads “How long does SEO take for a law firm?” tells the model exactly what that section answers. A paragraph buried in running text may contain the same information but requires the model to infer its purpose.

The two-layer model: depth first, technical signals second

Most guidance on AI search optimisation is written as if the problem is entirely technical. Fix your schema. Add FAQs. Use shorter paragraphs. These aren’t wrong — but they treat the symptom, not the cause. The correct model has two layers, and they must be built in sequence.

Layer 1 — Building content AI models treat as a credible source

This is the harder layer and the one that takes the most time. It requires:

Publishing substantive content on your topic. Not one or two overview articles, but a body of work that demonstrates genuine expertise across the full range of questions your clients ask. AI models assess topical authority — a site with twelve well-researched articles on content marketing for law firms will be cited over a site with two thin ones, all else being equal.

Demonstrating the expertise, not just claiming it. Named authors with stated credentials, inline citations of primary sources (regulatory bodies, research studies, official guidance), specific examples from practice. “In our experience working with specialist professional services firms” is more citable than “experts agree that.”

Building a content cluster, not isolated articles. A single article on a topic has one chance to earn a citation. A cluster of interconnected articles on the same topic — each linking to the others, each covering a different dimension — creates a semantic signal that the site is genuinely authoritative in that area. That signal compounds over time.

If this layer isn’t in place, Layer 2 has nothing to work with. Schema markup applied to thin content is still thin content.

Layer 2 — The technical signals that get credible content cited

Once you have substantive content, the technical layer determines how efficiently it gets picked up. This is where the formatting and schema work happens — and where most guides spend all their time.

How to optimise content for AI search — the practical steps

Step 1 — Answer the query directly in the opening paragraph

AI models assess whether content answers the user’s question. The opening paragraph should do exactly that, without preamble. State what the article covers, what the answer is at a high level, and why it matters. Content that makes the model work to find the answer is less likely to be cited than content that leads with it.

Step 2 — Structure content for extraction

Use clear H2 and H3 headings that describe what each section answers — ideally phrased as questions or as explicit topic labels. Use bullet points and numbered lists for any information that can be broken into distinct items. Keep paragraphs short: two to three sentences as a default.

The goal is to make each section function as a standalone answer that an AI model can extract without the surrounding context. If a section only makes sense when read in sequence after the previous three sections, it won’t be cited as a standalone source. Restructure until each section answers its own question independently.

Step 3 — Add inline authority citations — not a footer list

Research from Digital Applied (April 2026) found that inline citations — placed at the specific sentence where the claim appears — earn 2.1× more AI citation lift than the same citations placed in a dedicated sources section or footer list. This is a meaningful difference. Cite the source at the point where you make the claim: “According to Google Search Central’s guidance on AI optimisation (verified May 2026)…” rather than listing sources after the article.

For professional services firms, this means citing the FCA, SRA, Law Society, or relevant professional body at the specific point in the article where you reference a regulatory requirement or sector standard — not in a general disclaimer at the end.

Step 4 — Add the right schema markup

Three schema blocks are standard for any article targeting AI citation:

Article schema — identifies the content as an article, names the author, and provides publication and update dates. This signals authorship and content type to AI crawlers.

FAQPage schema — marks up the FAQ block at the end of the article. Every article that includes a FAQ block should have this schema implemented correctly.

BreadcrumbList schema — signals the content’s position within your site hierarchy to AI systems. A simple Home › Blog › [Article title] breadcrumb, marked up in JSON-LD, provides a navigational signal that contributes to AI citation rates.

All three are implemented in JSON-LD format in the <head> of the HTML file. According to Digital Applied’s 1,000 AI Overviews Citation Study (April 2026), schema-marked pages (Article + BreadcrumbList as a baseline) are cited 2.3× more often than otherwise-comparable unstructured pages after controlling for domain authority — making schema the single largest engineerable lever in their dataset of 4,200 cited URLs.

Step 5 — Build topical depth across a cluster, not isolated articles

A single well-optimised article on a topic creates a single citation opportunity. A cluster of five or six interconnected articles on the same topic — each answering a different question, each linking to the others — creates a topical authority signal. AI models evaluate not just whether a page answers a question well, but whether the site it lives on has genuine depth on the topic.

For a solicitor’s firm, this might mean articles covering content marketing for law firms, SEO for solicitors, SEO for personal injury law firms, and the case for specialist versus generalist agencies — all internally linked, all reinforcing each other. The first article establishes the entity; subsequent articles build the credibility layer that AI models use when deciding which source to cite for a given query.

Step 6 — Maintain content depth and accuracy, not just freshness

A finding that surprises many marketers: AI Overviews are not significantly recency-biased. Digital Applied’s 1,000 AIO Citation Study (April 2026) found that the median cited page is 14 months old, and page recency did not correlate with citation rate after controlling for domain authority and content structure. For most professional services content — which covers regulatory frameworks, service explanations, and sector analysis rather than breaking news — this is good news. A well-built article tends to remain citable over time.

What does warrant review is substantive accuracy. When regulatory guidance changes (FCA, SRA, Law Society), when market conditions shift materially, or when new research becomes available on a topic you’ve written about, update the relevant sections with genuinely new content — not just a refreshed date. The goal is that every factual claim in the article remains accurate and, where primary sources are cited, those sources are still live and unchanged. That’s different from “keep publishing to stay fresh” — it’s maintaining the depth and accuracy that made the article worth citing in the first place.

What this looks like in practice for a professional services firm

Why the content depth problem is harder for regulated firms

Professional services firms face a specific challenge that generic AI optimisation guides don’t acknowledge. In regulated sectors — financial services, legal, surveying — the expertise that would make the best content is often either compliance-constrained or simply unpublished. Firms worry about making claims they can’t support, advice that could be misconstrued as legally binding, or specificity that creates liability.

The result is that most professional services content is written to a standard of vagueness that avoids risk but also avoids depth. This content won’t earn AI citations because it has nothing distinctive to offer — AI models can find equally vague general information everywhere.

The correct approach isn’t to abandon caution — it’s to be specific about what the firm genuinely knows while being accurate about the boundaries of that knowledge. “In our experience working with specialist financial services clients, the content that tends to generate enquiries is…” is specific, expert, and defensible. It gives AI models something to cite that they can’t find in a generic overview.

Authority signals in practice: named sources, inline citations, verified dates

For a professional services content article to earn AI citations, every factual claim should be supported at the point where it appears. This means:

• Regulatory references cited by the correct document and body — not “industry guidance says” but “FCA COBS 4.2 requires…” or “SRA Code of Conduct 8.8 states…”
• Research claims cited with source and date — not “studies show” but “according to [specific research], verified [month/year]”
• Market claims qualified appropriately — not “content marketing works for law firms” but “in our experience with specialist professional services clients, systematic content production tends to…”

This level of precision is harder to maintain than writing confident-sounding generalities. It is also exactly what AI models are looking for when they decide whether your content is worth citing.

Frequently asked questions