SEO Content Optimisation: EEAT, QRIES and Topical Authority in the Age of AI Search

In one line: stop writing pages that target keywords and start building bodies of work that make you the obvious source on a topic, for both a search engine and a language model.

Most advice about optimising content still describes a world that no longer exists. It tells you to pick a keyword, put it in the H1, repeat it a handful of times and add a few internal links. That world ended somewhere around 2021. A single page now ranks for thousands of related queries. Google shows an AI summary on roughly 30% of searches. And a growing share of the people you want to reach never see a results page at all, because they asked ChatGPT, Gemini, Perplexity or Claude instead.

This guide replaces the advice we published in our earlier EEAT article. The framework is the same in spirit but far more useful in practice. It has three layers:

• EEAT is the principle. It tells you what "good" means.
• QRIES is the technique. It tells you how to actually produce work that meets that standard, sentence by sentence.
• The trifecta of topical authority, entities and structure is the machine layer. It tells you how to make that quality legible to systems that read rather than browse.

Get all three right and you are not optimising for Google or for ChatGPT. You are optimising for being genuinely the best answer, then removing every obstacle between that answer and the machine trying to find it.

Key takeaways

• Trust is the most important part of EEAT. Google says so explicitly. Experience, expertise and authoritativeness exist to build it.
• QRIES gives you a repeatable production standard: Quotes, Research, Images, Examples, Statistics. If a draft has none of these, it is a summary of the internet, not a contribution to it.
• Topical authority measurably shortens time to visibility. In a Graphite study of 332 URLs across 12 domains, articles published on high topical authority sites earned their first impression and first click significantly faster than those on low authority sites.
• Keyword targeting and topical authority are different games. One asks "what page do I need?" The other asks "what do I need to become known for?"
• LLMs mostly do not recall you from training. They retrieve you. ChatGPT, Gemini and Perplexity run a search, then summarise the results. Classic search visibility is the entry ticket, and being mentioned inside other people's cited pages is the multiplier.
• Ignore most of the AEO folklore. Google has publicly stated that llms.txt files, artificial content chunking and manufactured brand mentions do not help visibility in its AI features.
• Content design sits between strategy and copywriting. It is the discipline that decides what a reader needs, in what order, in what format. It is not just words.

Part 1: EEAT is the principle, not the checklist

EEAT stands for Experience, Expertise, Authoritativeness and Trust. It is not a ranking factor. Google is unambiguous about this: EEAT is a concept its quality raters use, and its automated systems use a mixture of signals that correlate with it.

That distinction matters enormously for a writer. You cannot "add EEAT" to a page the way you add a meta description. You can only produce work that a reasonable person would recognise as trustworthy, and then make the evidence of that trustworthiness visible.

What each letter actually asks of you

Google states that of the four, trust is most important, and that the others contribute to trust. Content does not have to demonstrate all four. A first hand product review can be helpful mainly on the strength of experience. A technical explainer can be helpful mainly on expertise.

The Who, How and Why test

Google recommends three questions as a self audit. They are unusually practical, so use them literally on your next draft.

Who created this? Is the author self evident? Is there a byline where a reader would expect one? Does that byline lead somewhere with real background?

How was it created? If you tested five products, say how you tested them. If you used AI assistance, be open about where and why. If the data is yours, describe the method and sample size.

Why does it exist? This is the one that catches people out. If the honest answer is "because this keyword has volume", the piece is unlikely to survive a core update. If the answer is "because our readers keep asking us this and nobody has answered it properly", you are on solid ground.

A five minute EEAT pass for any draft

Run this before you file. Each line is either present or it is not.

1. Named author with a link to a page proving relevant background.
2. A visible publication date and, if updated, a genuine change log rather than a date refresh.
3. At least one thing on this page that exists nowhere else: original data, an original photograph, a first hand account, a client example.
4. Every factual claim traceable to a named source with a link.
5. Trade offs and limitations stated openly, not buried.
6. No claim that overstates what the evidence supports.
7. A clear answer to "why should this publisher be the one telling me this?"

Part 2: QRIES is how you actually write it

EEAT tells you what good looks like. QRIES, a framework used by Semrush's content team, tells you how to produce it consistently. It stands for Quotes, Research, Images, Examples, Statistics.

Its real value is that it converts an abstract quality standard into five things you can look for in a draft and count.

Q for Quotes

Why it works: a quote from a named practitioner does something you cannot do alone. It shows you went and found someone who knows, and it hands the reader a second voice to weigh your argument against.

How to do it:

• Keep a running expert database organised by topic. Name, role, where they publish, how to contact them, and quotes you have already gathered.
• Source from industry publications, conference talks, podcast transcripts, LinkedIn posts and expert platforms.
• Attribute properly every time: name, role, organisation, and a link to the original.
• Prefer a specific, arguable quote over a bland one. "Most brands under invest in this" is filler. A quote that stakes out a position is evidence.

Writer's rule: one genuine expert voice per substantial article, minimum.

R for Research

Why it works: research is what separates an article that adds to a topic from one that recycles it. Google explicitly warns against content that mainly summarises what others have said.

How to do it:

• Split your research into fast and slow. Fast research is a wide scan to map the territory and identify which sources are worth your time. Slow research is deep reading of the two or three sources that actually contain something new.
• Apply an 80/20 split. Roughly 20% of what you read produces 80% of what makes the piece distinctive. Find that 20% early.
• Diversify formats deliberately. Blog posts are the most crowded and least original source. Conference videos, primary research papers, earnings calls, court filings, product documentation and interviews all contain material your competitors have not read.

Writer's rule: if every source in your reference list is a blog post from a competitor, start again.

I for Images

Why it works: visuals are functional, not decorative. A diagram can carry a relationship that would take three paragraphs to describe badly. Google's own guidance on generative AI features calls out high quality images and video as an opportunity to appear beyond web page links.

How to do it:

• Use charts, diagrams, annotated screenshots and custom graphics. Not stock photography of people pointing at laptops.
• Place visuals at the points where a reader is most likely to lose the thread: after a definition, inside a step by step, alongside a comparison.
• Write descriptive alt text that would make sense to someone who cannot see the image. This is an accessibility requirement first and a machine readability benefit second.
• You do not need a designer. Tools like Canva and Figma will produce a clean diagram in twenty minutes.

Writer's rule: one original visual per major section.

E for Examples

Why it works: examples make abstract concepts concrete, and they are the single fastest way to demonstrate experience. An example is proof that you have seen the thing happen.

How to do it:

• Draw first on your own work, your clients and your own mistakes. These are the examples nobody else has.
• Use the full range: case studies, hypothetical scenarios, analogies and short narratives.
• When researching, add "case study" and "results" to your queries to surface documented examples rather than opinion.
• One well developed example beats four thin ones.

Writer's rule: every claim that a reader might reasonably doubt gets an example attached to it.

S for Statistics

Why it works: statistics turn assertions into verifiable facts. They also attract links, because other writers need numbers and will cite the page that gave them one.

How to do it:

• Build a statistics database for your industry. Number, source, date, methodology and a link. Refresh it quarterly.
• Always give context. "AI search traffic rose 527% year on year" means little without knowing it moved from roughly 17,000 to 107,000 sessions across the properties studied.
• Cite the original study, not the blog that cited the study. Chasing a statistic back to its source frequently reveals that it does not say what everyone claims.
• Visualise the important ones.

Writer's rule: no unsourced number, ever.

QRIES mapped to EEAT

Part 3: Topical authority beats keyword targeting

This is the most consequential shift in the last five years, and the one most content teams have not yet made.

The difference, stated plainly

Keyword targeting asks: which query should this page rank for? It produces a library of individually optimised pages, each competing on its own merits, each starting from zero.

Topical authority asks: what subject do we want to be known for, and what is the complete set of things a person needs to understand about it? It produces a connected body of work where each new piece inherits credibility from everything around it.

The mechanism is not mystical. Search engines now understand topics, not just strings. Because they understand topics, they can assess how relevant a domain is to a topic. Google publicly confirmed in 2023 that topic authority exists and influences results.

A worked example

Say you sell project management software.

The keyword approach produces: "project management software", "project management software for small business", "best project management software 2026", "project management software pricing". Four pages, four battles, four sets of competitors, and considerable overlap that risks cannibalising your own rankings.

The topical authority approach starts by mapping the subject. What does someone evaluating project management software actually need to know? Methodologies. Team size thresholds. Integration requirements. Migration risk. Pricing models. Failure modes. Who should not buy this category at all.

You then build a small number of substantial pages, each covering a cluster of intent rather than a single query, connected by internal links that express how the ideas relate. The commercial page inherits authority from the explainers. The explainers inherit authority from the original research. A new article on the same subject enters the ranking pool with the site's existing topical credibility already attached to it.

The evidence

Graphite, a growth agency with a research practice, studied 332 URLs published across 12 domains in June and July 2023, spanning technology, software, healthcare, e-learning, financial services, manufacturing and media. They split the URLs into three buckets by topical authority score and tracked how quickly each earned its first impression and first click.

The finding: high topical authority significantly decreases time to visibility. Pages in the highest bucket earned their first impression and first click faster, a greater share of them achieved visibility within the first three weeks, and they reached impression milestones sooner.

This is the practical argument that wins budget. The most common objection to content investment is that it takes too long to see results. Topical authority is the lever that shortens that timeline.

Graphite prioritises content using three factors together: topical authority, topic search volume and business value. Any one on its own misleads you.

How to build topical authority, step by step

1. Choose a defensible territory. Not "marketing". Something narrow enough that you could plausibly become the best source on the internet within twelve months.
2. Map the topic properly. List every question a person moves through from unaware to decided. Group them by intent, not by keyword. Each cluster of questions sharing one intent becomes one page, not five.
3. Build depth before breadth. Twenty excellent pages on one subject outperform two hundred thin pages across twenty subjects. Google's spam policies explicitly target scaled content produced across many topics in the hope some of it performs.
4. Link the cluster deliberately. Internal linking is one of the few technical interventions with tested impact. In a Graphite A/B test across 199 URLs, the group that received systematic internal linking saw a 42.4% increase in average daily impressions within 50 days, compared with a 0.7% increase for the control group. The effect was largest on pages that previously had little traffic.
5. Publish consistently in that territory. Authority compounds. Sporadic publishing does not.
6. Prune ruthlessly. Pages that are thin, outdated or off topic dilute the signal. Improve them or remove them.

If you would rather have this run for you than run it yourself, that is exactly what our answer engine optimisation and fractional CMO engagements are built to do.

Part 4: The machine layer: entities, schema and structure

Everything above makes content good. This section makes it legible.

Entities: be a thing, not a string

An entity is a distinct, identifiable thing that a knowledge system recognises: a company, a person, a product, a place, a concept. Search engines and language models reason about entities and the relationships between them, not just about words.

Being recognised as an entity is what allows a system to associate your brand with a category. Practically, that means:

• Be consistent. Same brand name, same spelling, same descriptor everywhere. Inconsistency splits the signal.
• Pair your brand with your category phrase. Content and mentions that place your brand next to your product category build the association you want a model to make.
• Be present where your category is discussed. Industry publications, review platforms, community forums, expert roundups. Models retrieve from these sources.
• Give machines an unambiguous description of who you are. A thorough About page, an Organization schema block, consistent profiles on the platforms that matter in your sector.

A word of caution here, and it comes directly from Google: pursuing inauthentic mentions is not as helpful as it sounds. The systems that reward high quality content and the systems that block spam both feed generative AI features. Earn mentions by being worth mentioning.

Schema: useful, but not the miracle it is sold as

Structured data helps machines parse what your page contains. Use it. But be honest about what it does.

Worth implementing:

• Article with author, datePublished and dateModified
• Organization on your site, with sameAs links to your verified profiles
• Person for author pages, with credentials and affiliations
• FAQPage where you genuinely have a question and answer section
• HowTo for genuine step by step processes
• BreadcrumbList for site hierarchy

The honest caveat: Google has stated plainly that structured data is not required for generative AI search and that there is no special schema you need to add for it. It remains a good idea as part of overall SEO because it makes you eligible for rich results. Treat schema as hygiene, not strategy.

Structure: write passages that survive being extracted

This is where content design and AI visibility converge. AI systems retrieve and summarise passages. A passage that only makes sense in the context of the three paragraphs above it is a passage that cannot be used.

Practical rules:

• Answer first, elaborate second. Open each section with a direct, self contained answer, then expand.
• Make headings match how people ask. "How long does topical authority take to build?" is more retrievable than "Timelines".
• Write self sufficient sentences. Avoid "as mentioned above" and "this approach" where a reader arriving cold would not know what you mean.
• Use tables for comparisons and lists for sequences. These structures survive extraction intact.
• Front load the important material. Readers scan in an F shaped pattern, concentrating attention on the opening lines and the left edge. Roughly 70% of users read only the first third of an AI Overview.

Note one more thing Google has said explicitly: you do not need to artificially chunk your content into tiny pieces for AI to understand it. Its systems handle multiple topics on a page. Write clear, well organised pages for humans. Clarity is the optimisation.

Content design: the discipline between strategy and copywriting

Content design is a user centred approach to creating content that solves audience needs through clear, accessible, evidence based communication. It sits between content strategy, which is the guiding foundation, and copywriting, which is one output of it.

It is not just words. Layout, sequence, format, chunking, labelling and what you choose to leave out are all content design decisions.

The process, adapted for a content team:

1. Research and discovery. Understand the users, their needs and the context in which they meet your content.
2. Define user needs. Write them as user stories: as a role, I want to do something, so that I get an outcome.
3. Channel and journey mapping. Where does this content sit relative to everything else the reader will encounter?
4. Content planning. What gets made, by whom, and how is it maintained?
5. Creation. Clear, concise, accessible, tailored.
6. Testing and feedback. Validate with real users, not internal opinion.
7. Iteration. Improve based on what testing showed.
8. Publishing and management. Deploy properly and maintain over time.

Good content, in this framing, does four things: it informs, it encourages an action, it builds trust, and it is consistent with the brand. If a page does not do all four, the problem is usually design rather than prose.

As Pascal put it, "if I had more time, I would have written a shorter letter." Editing is content design.

Part 5: What actually influences visibility in LLMs

This is the part most guides get wrong, usually by treating language models as mystical.

They retrieve far more than they recall

Early language models were trained on static snapshots of the web, often more than a year old. Optimising for a system with a twelve month feedback loop was close to impossible.

That is no longer how most AI answers work. ChatGPT, Gemini, Perplexity and Google's AI Overviews now use retrieval augmented generation. When you ask a question, the system runs one or more searches, retrieves current pages, and summarises them. Google's own documentation confirms that its generative AI features are rooted in its core Search ranking and quality systems, and uses query fan out, generating several related queries alongside your original one.

Three consequences follow, and they are the whole strategy:

1. Classic search visibility is the entry ticket. If you are not retrievable, you cannot be cited. A page must be indexed and eligible to appear with a snippet before it can appear in Google's AI features at all.
2. Query fan out rewards topical depth. The model does not just search your target query. It searches around it. A site that comprehensively covers the neighbouring questions gets retrieved more often, which is exactly what topical authority produces.
3. You need to be in other people's pages, not just your own. For a general question like "what is the best project management software", the retrieved set is mostly publisher round ups, not vendor sites. The more specific the question, the more an owned strategy works. The more general the question, the more an earned strategy works.

Why this traffic is worth chasing even when volume is low

The volume argument against AI search is weakening fast, and the value argument was never weak.

• AI search traffic rose roughly 527% year on year in one 2025 analysis of 19 GA4 properties, moving from about 17,000 to 107,000 sessions.
• Semrush estimates the average AI search visitor is worth around 4.4 times more than a traditional organic search visitor on a conversion basis.
• About half of the links ChatGPT cites point to business and service websites, with news and media at 9.5% and blogs at 8.3%. Company sites with genuinely useful content have a real chance of being cited.
• ChatGPT users click out to external sites roughly 1.4 links per visit, compared with about 0.6 for Google users.

Meanwhile, the panic is overstated. Graphite and Similarweb analysed more than 40,000 of the largest US sites and found organic search traffic down 2.5% year on year, not the 25% to 50% collapse widely reported. AI Overviews do reduce click through rate by about 35% when they appear, but they appear on roughly 30% of searches, mostly informational ones, and they largely replaced featured snippets, which also suppressed clicks.

The sensible reading: search is not dying, AI search is growing, and the work that wins in one wins in the other.

What to ignore

Google has published direct guidance on this, which saves everyone a great deal of wasted effort:

• llms.txt files and similar "special" markup. Google Search does not use them. Maintaining one neither helps nor harms.
• Artificially chunking content. Not required.
• Rewriting content specifically for AI systems. They understand synonyms and meaning. You do not need every long tail variation.
• Manufactured brand mentions. Inauthentic mentions are not as helpful as they appear, and spam systems feed the same pipeline.
• Over indexing on structured data. Useful for rich results, not required for AI features.
• Technical audits as a substitute for content work. Technical SEO produces a satisfying volume of charts and tasks and, internal linking and rendering aside, rarely drives significant impact. Expect the same pattern to repeat with "technical AEO".

How to measure it

The old metrics do not describe this landscape. Track these instead.

Two practical notes on measuring AI visibility. First, the same prompt produces different answers on different runs, so you need a sample. Graphite's research found that ten responses per prompt is enough for a reasonable position estimate, with a mean absolute error of about one position. Second, chase visibility before position. There is no point optimising your average rank in answers you rarely appear in at all.

Part 6: The 30 day plan

If you write content and want to act on all of this next Monday, here is the sequence.

Week 1: Audit

• Pick your single most important topic. Just one.
• List every page you already have on it. Score each against the seven point EEAT pass and the QRIES scorecard.
• Test ten priority questions in ChatGPT, Gemini, Perplexity and Google. Record whether you are cited, whether competitors are, and which URLs are cited.

Week 2: Fix what exists

• Take your three strongest pages on the topic and rewrite the opening of every section so it answers directly.
• Add what QRIES says is missing: one expert quote, one original visual, one worked example, sourced statistics.
• Add or fix the author byline and link it to a page with genuine credentials.
• Add Article, Person and Organization schema.

Week 3: Connect and fill

• Map the internal links across your cluster. Every page should link to the pillar and to at least two siblings, with descriptive anchor text.
• Identify the two biggest gaps in the topic map and brief them properly.
• Identify the three most frequently cited third party URLs for your priority questions. Work out how to be legitimately mentioned in them: original data worth citing, expert contribution, a genuinely newsworthy piece of research.

Week 4: Publish and set up measurement

• Publish the highest value gap piece, built to standard from the start.
• Set up tracking for AI visibility on your ten priority questions, ten runs each.
• Set a baseline for organic traffic by cluster so you can see the compounding effect later.

Then repeat monthly on the same topic until you are the obvious source. Only then move to the next topic. If you want a second pair of hands on that sequence, get in touch or take a look at our services.

Frequently asked questions

Is EEAT a ranking factor?

No. Google states that EEAT is not itself a specific ranking factor. Its automated systems use a mixture of signals that identify content demonstrating experience, expertise, authoritativeness and trust. Those signals carry more weight for topics that could significantly affect health, financial stability, safety or wellbeing, known as Your Money or Your Life topics.

What is the difference between keyword targeting and topical authority?

Keyword targeting builds one page per query and each page competes alone. Topical authority builds a connected body of work covering a subject completely, so each new page inherits credibility from the existing cluster. Research from Graphite across 332 URLs found that pages on high topical authority sites reach their first impression and first click significantly faster than those on low authority sites.

How long does it take to build topical authority?

There is no fixed timeline, but the mechanism is compounding, so the curve is slow then steep. The practical evidence is that high topical authority shortens time to visibility for each new piece, which means the seventh article in a cluster performs much faster than the first. Plan in quarters, not weeks.

Do I need schema markup to appear in AI Overviews or ChatGPT?

No. Google has stated that structured data is not required for generative AI search and that no special schema markup exists for it. Implement schema anyway, because it makes you eligible for rich results in traditional search, but do not treat it as an AI visibility strategy.

Does AI search actually reduce my traffic?

Less than the headlines suggest. An analysis of more than 40,000 large US sites found organic search traffic down 2.5% year on year. AI Overviews reduce click through rate by roughly 35% where they appear, but they appear on around 30% of searches, concentrated on informational queries. Commercially valuable queries are far less affected.

What is QRIES?

QRIES stands for Quotes, Research, Images, Examples and Statistics. It is a production standard used to make content quality repeatable, ensuring every substantial piece contains expert voices, genuine research, functional visuals, concrete examples and sourced data.

What is content design and how is it different from copywriting?

Content design is a user centred approach to creating content that solves audience needs through clear, accessible, evidence based communication. It sits between content strategy, which is the foundation, and copywriting, which is one of its outputs. Content design decides what a user needs, in what order and in what format. Copywriting is the words that result.

Sources

• Creating helpful, reliable, people-first content, Google Search Central
• Optimizing your website for generative AI features on Google Search, Google Search Central
• Study Shows That High Topical Authority Leads to Faster Organic Search Visibility, Graphite, April 2024
• AB Tests Show That the Internal Links API Increased Daily Impressions by 42% After 50 Days, Graphite, April 2024
• Ultimate Guide to AEO, Graphite, April 2025
• Debunking The Myth That Search Is Dying, Graphite and Similarweb, January 2026
• How to Track Entity Position in AI, Graphite, May 2026
• 26 AI SEO Statistics for 2026, Semrush
• The Content Trifecta: How Strategy, Design and Writing Drive Rankings, Semrush Marketing Meetup, Barcelona. Sessions by Carlos Silva (Content Writing in the Post-Keyword Era), Teresa Lee (Content Design: A Primer for Marketers) and Kate Busby (SEO Strategy Shift).