Content marketing for SEO: how to build one semantic core for your site

One semantic core for search

A semantic core is one shared content map of the topics and entities your brand wants to own. Every page you publish, for Google or for AI answers, comes from that map. Teams that keep separate keyword lists for the blog and product pages end up with pages that compete for the same intent when they add a third list for "AI marketing strategy." Google doesn't split the work either. Danny Sullivan, who was Google's Search Liaison, told a WordCamp US audience in September 2025 that "good SEO is good GEO", where GEO means generative engine optimization.

In B2B content marketing, a semantic core has layers. Take an e-signature vendor doing SaaS content marketing. Its core has four parts, and each needs its own pages:

• The category entity: electronic signatures, and how they differ from scanned or typed names.
• Problem entities: contract turnaround time and compliance with the EU's eIDAS regulation on electronic identification and authentication, which also covers trust services.
• Buyer roles: the legal counsel who checks validity and the sales operations lead who wants faster deals.
• Decision terms: pricing models and integrations with the CRM. Comparisons against other ways of signing also belong here.

The size of the core depends on how broad the category is. In startup content marketing, a single-product vendor in a narrow niche covers its core with fewer pillar pages than a platform that sells to both finance and HR. Use a content strategy template to map every existing or planned URL to exactly one topic in the core. When two pages chase the same intent, merge them before you write anything new.

Why AI needs semantic depth

AI assistants break each question into smaller questions and pull a passage for every one. A site that only covers the main term gives them little to cite. Google described its AI Mode as "breaking down your question into subtopics" and running a large batch of searches at once on the user's behalf. Those sub-queries reach far down the results. In a March 2026 Ahrefs analysis of 4 million AI Overview URLs, only 38% of AI Overview citations came from pages ranking in Google's top 10 for the same keyword.

Ahrefs found that an AI Overview on the results page goes with a 58% lower clickthrough rate for the top-ranking page, on December 2025 data. Seer Interactive studied 3,119 informational queries and measured a 61% drop in organic CTR on queries that showed an AI Overview. If ranking alone used to protect your traffic, it doesn't anymore.

To build a content marketing strategy that AI assistants cite, cover each sub-question a model generates. Give each one a self-contained answer it can lift without the paragraphs around it. Then back the answer with evidence. In the original generative engine optimization study by Princeton researchers, adding statistics and quotations to a page lifted its visibility in AI answers by up to about 40%. Keyword stuffing made the same pages roughly 10% less visible. Depth here means the specific claims a model can check, and padding a page to a word count doesn't add any.

Building your topic cluster map

For AI search visibility, a topic cluster has one pillar page per core topic. Around it sit cluster pages that each answer one sub-question and link back to the pillar. The pillar defines the topic and briefly answers each sub-question, while each cluster page explores one sub-question in depth. This is the structure we build for clients, and you can follow it in order:

1. Choose the pillars from your semantic core. One pillar per core topic. If a candidate pillar can't hold at least a few distinct sub-questions, make it a cluster page instead.
2. Collect the real sub-questions. Pull them from your Google Search Console queries and from the questions buyers ask on sales calls. Then run the main topic through ChatGPT and note which follow-up angles it brings up.
3. Give each intent one page. Questions that need their own steps or figures get cluster pages, as do questions that need a comparison. A question with a two-sentence answer stays as a paragraph on the pillar.
4. Wire the internal links. Every cluster page links to its pillar, and the pillar links to every cluster page. Sibling pages link to each other only where a reader would go next.
5. Put the answer first on every page. Kevin Indig, a growth advisor, looked at 18,012 verified ChatGPT citations and found that 44.2% of citations came from the first 30% of a document.
6. Give every page an owner and a review date. Clusters suffer content decay one page at a time, and someone has to own each page.

A company with a complex product has deep clusters under fewer pillars, while a services firm has broader, shallower clusters. Test the finished content map against two rules (what a content map is defines the term). Give each cluster page a question distinct from its pillar's. Keep the pillar within one click of every cluster page.

Keyword research meets entity coverage

Keyword research identifies the phrases people type, while entity coverage identifies the things a model expects to find connected to your topic. A content plan built on only one of them leaves gaps. An entity is a named thing with a fixed meaning, such as a product category or a job title; standards and regulations are entities too. Language models learn which brands belong with which entities by seeing them mentioned together across the web. A page that ranks for "contract signing software" but never names the regulation that makes signatures valid has given the model half the picture.

Run the two together, cluster by cluster. For each cluster, list the keywords with their search volume and intent. Then read the pages and AI answers that already cover the topic, and write down every entity they mention. A keyword with no entities behind it is a phrase without a real subject. An entity that no keyword points to is still worth a paragraph, because models expect to see it even when people rarely type it.

Use consistent names for your product and category across your site and third-party profiles. Keep your founders' names consistent there too, and use the same names in directories and press coverage. Inconsistent naming splits the signal across two entities. Google says AI Overviews and AI Mode have no special schema requirement, so clear writing and consistent naming do most of the work.

Content types AI cites

Listicles and long-form articles earn the largest share of AI citations, and product pages come next. The best mix for you depends on the intent behind each query. Wix Studio's AI Search Lab studied 75,000 AI answers and more than 1 million citations in a study of ChatGPT and Google AI Mode that also covered Perplexity. It found that listicles led at 21.9% of citations.

Listicles and comparisons for commercial queries

Ranked lists and comparisons win commercial queries, where the buyer is weighing options. In the Wix data, listicles took 40% of commercial-intent citations, nearly double any other format. HubSpot's State of AEO 2026, where AEO means answer engine optimization, found that comparison content earns a 95% citation rate in ChatGPT, the highest of any format on any engine it measured. A list that ranks its publisher first with no stated criteria reads as an advert to Google and to readers. State the criteria and score every option on them, your own included.

Long-form articles for informational queries

Explanatory articles win when the reader wants to understand something before buying. Wix found that articles were cited 2.7 times more than other formats on informational queries. ChatGPT leans the hardest this way, and 43% of its citations come from articles and listicles combined. Your pillar pages belong here. They work best when each section answers one sub-question in its first sentence.

Product and category pages for buying queries

Product pages get cited when the query names a specific need or a spec, and price queries draw citations too. Pages with clear specs and stated prices give a model facts it can quote word for word. They need the same answer-first writing as an article. 3DPrinterOS, a 3D printing management platform we work with, achieved #1 visibility in ChatGPT for industry-specific queries and turned AI search into a steady source of 3–4 qualified inbound leads/day.

Community discussions for Perplexity

Perplexity draws heavily on forums. In the same Wix study, 17% of its citations came from discussions on Reddit and similar sites. You can answer questions in those threads under your real name and make sure your own pages say the same things you say there.

Measuring search visibility

You measure search visibility on two scoreboards. The two measures are classic rankings and clicks from Google Search Console and your brand's presence inside AI answers. Search Console tells you whether Google can find and rank your pages, while AI answers tell you whether a model mentions you when a buyer asks. A brand can rank well and still be missing from the AI answer that sits above the results.

AI visibility is measured with a prompt set, a fixed list of buyer questions you run on a schedule. With weekly prompt sets, our tracking measures mentions and share of voice in ChatGPT and Google AI Overviews, where most AI search traffic happens. It also records sentiment and citations. Weekly runs matter because the answers keep changing. Ahrefs found a 70% chance an AI Overview changes between observations, so a single check is stale within days. Run consistently, a prompt set shows a trend. Snoika's founder Anton Vedeshin reported to Startup Reporter 1StopVAT's move from 2.7% to 28.3% ChatGPT visibility in 16 weeks.

Connect the two scoreboards to revenue. Watch whether citation gains come before rises in branded search and in conversions tracked in Google Analytics 4 (GA4). Seer Interactive found that brands cited in AI Overviews earn 35% more organic clicks than brands that aren't cited on the same queries. Check that your prompt set uses the words buyers actually use. A prompt list written in your internal jargon will tell you you're doing fine.

Scaling content without losing authority

You can scale content safely when each new page answers a real question your site hasn't answered yet, and a person checks it before it goes live. Google's line is about why a page exists, and it doesn't care how the page was made. Its March 2024 spam policy defines scaled content abuse as when many pages are generated "for the primary purpose of manipulating Search rankings and not helping users." The policy applies "no matter how it's created." Automation is allowed. The trouble starts when you produce pages with nothing original in them.

Volume holds up when your cluster map limits what gets written and a review step catches errors before readers do. Our team runs automated workflows with a human review layer on top. Where the pages live matters too, because authority builds on the domain that hosts them. We use a DNS connection to publish on the client's own domain. This works on any CMS and requires no migration. The client reviews and approves each page in the dashboard before it goes live. Agrello, an e-signature company, recorded +125.6% organic search sessions YoY and generated 51.8K impressions from Snoika-created pages.

Translation causes problems as content scales. A word-for-word translation keeps the source market's questions and leaves out the ones local buyers ask. We create content in 135+ languages and regional variants, and every language version still needs its own check of sub-questions and entities. A German buyer searching for e-signature validity cares about different rules than a buyer in Texas.

Auditing existing content for gaps

A content audit compares what your site already covers with your semantic core and the prompts buyers ask. The audit identifies pages to keep or update and decides which of the remaining pages to merge or cut. Most sites get more from fixing old pages than from writing new ones, because the old pages already have links and history. Run it in this order:

1. Export your pages and queries. Pull every indexed URL with its queries from Google Search Console, and conversions per page from GA4.
2. Map each URL to one core topic. Pages that fit nowhere are candidates to cut. Merge two pages that cover the same topic and intent.
3. Flag stale pages. In a study of 17 million citations, Ahrefs found that AI-cited content is 25.7% fresher than Google's organic results. Update the facts and dates on pages that carry your core topics.
4. Run your prompt set. Note where competitors appear in AI answers and you don't. Each of those prompts points to a missing sub-question or a page that buries its answer.
5. Write the gap list. Missing sub-questions become new cluster pages. Buried answers get moved to the top of the page they're on.

Access decides how deep a content gap analysis can go. When we start work with a client, onboarding covers their category and buyers, along with the goals they want to achieve. We ask for access to GA4 and Google Search Console. We also request Google Tag Manager access, plus the CMS when pages will be published. Technical problems show up at this stage. ABS Technologies saw 6x SERP impressions and improved keyword indexability within the first 30 days, when its search visibility also increased significantly.

Ways to run SEO content

You have four common ways to run this work. They differ in cost and in who writes and publishes the pages. Coverage of AI visibility tracking varies too. The table compares them on the same attributes, with prices from 2025 and 2026 sources.

Digital Applied's 2026 pricing review puts content marketing retainers at $3,000 to $6,000 monthly for boutique agencies and at $6,000 to $12,000 for mid-size firms. GoodFirms reports that agencies now charge AI search optimization separately, at an average of over $900 per month. An in-house hire costs more than salary alone, and Glassdoor puts the average senior content manager at $139,924 per year before tools and benefits.

Our plans are priced weekly, which makes them easy to underestimate. Snoika's Starter plan runs one strategy at a time for €124 a week, with weekly tracking and a monthly AI visibility report. If you already have writers and only need to know where you stand, a monitoring-only tool costs less. If you want tracking and publishing handled by one team that already knows your gap list, an all-in-one platform saves you the work of keeping two vendors in step.

What to do next

Run the audit before you buy anything. The gap list it produces tells you whether you're short of writers or short of measurement, or both, and that answer decides which option in the table fits you. Track the prompts you lose as closely as the ones you win, because a competitor's new page shows up there weeks before it shows up in your traffic.

Content marketing for SEO is measured by how much of your core you cover over time. Our team uses site audits and visibility tracking in Google and AI assistants to identify missing pages, which we then publish on your domain. To start with your own gap list, book a call with us.