Links, mentions and AI search visibility

0:00 Links, mentions and AI search

Ask an AI assistant for the best accountant in your city, or the best project management tool for a small agency. You'll get a confident paragraph and a few cited sources. So here's the question every marketing lead is now asking. How do we get into that answer? And do links still matter if people don't click? In this lecture you'll learn how links and brand mentions feed AI search, what Google explicitly tells you not to do, and how to measure your visibility without fooling yourself.

0:38 Why it matters

Why does this matter? Because a whole industry has appeared overnight selling AI citations and LLM mentions. Some of it is sensible PR. A lot of it is the old link spam wearing a new outfit. If you understand how these systems actually choose sources, you can put your budget into things that work in both classic search and AI answers, and avoid tactics that Google has now explicitly put inside its spam policies.

1:10 The key idea

Here's the key idea. For Google, AI Overviews and AI Mode are built on the same crawling, indexing and ranking systems as regular Search. Google's own guide, published in May twenty twenty-six, says optimising for these features is essentially still SEO. Think of it like a restaurant critic writing a round-up of the best places in town. The critic doesn't invent new restaurants. They draw from places they already know and trust. AI answers draw from pages the search system already knows and trusts.

1:47 Query fan-out

There's one mechanism worth understanding. Google calls it query fan-out. For a complex question, the model runs several related searches at the same time, then combines the results into one answer. So if someone asks which HR software suits a fifty person company in Riyadh, the system might search for payroll features, Saudi labour law compliance, pricing, and reviews, all at once. Your page doesn't need to match the exact question. A deep, useful page can be cited because it answers one of those sub-questions really well.

2:25 What feeds AI answers

So where do links and mentions fit? Relevant editorial links still help pages get discovered and rank, and pages that rank well are more likely to be retrieved. Unlinked mentions in reputable places matter more than before, because AI answers often summarise what the web says about a brand. Reviews, forums and videos get summarised too. And consistent entity information, your official name, description and profiles, helps systems attribute facts to the right company. Being genuinely referenced, with or without a hyperlink, now shows up in more places.

3:03 Google says don't

Now, what Google says not to do. Its guide calls out inauthentic mentions, meaning seeding your brand across low quality blogs and forums to influence answers. And since May twenty twenty-six, Google's spam documentation says that trying to manipulate generative AI responses in Search falls under its spam policies. It also says you can ignore llms dot txt files for Google Search, content chunking and AI style rewriting. So the guaranteed AI citations package is not a clever new channel. It's a risk.

3:39 Example 1 (simple)

Worked example one, simple. A freelance wedding photographer in Lahore notices that AI assistants recommend three other photographers when asked about wedding photographers in the city. The sources cited are a local wedding directory, a bridal magazine's feature, and a forum thread. What does she do? She makes sure her directory profile is complete and accurate, pitches a story to the bridal magazine about a winter wedding trend she's documented, and politely asks recent couples to leave honest reviews. No shortcuts, just showing up where the answers come from.

4:18 Example 2 (illustrative)

Worked example two, realistic and illustrative. A Riyadh software company, call it Qimma HR, sees that answers to best HR software for Saudi SMEs cite two comparison articles and a Reddit thread. Qimma isn't mentioned. The sales director wants to buy placements. Instead, the team publishes an annual survey of Saudi SME HR practices with a transparent methodology. They pitch the findings to regional business media and HR newsletters, and offer their head of compliance as a source on labour law changes.

4:54 Outcome (illustrative)

They also ask new customers for honest reviews on two respected software review platforms after onboarding, with no incentives. Six months later, and again these are illustrative results, they've earned coverage in several business publications, some linked and some not. Branded searches have risen. And in their monthly prompt panel, Qimma now appears in some answers, cited via the survey coverage. The lesson is that the goal was never an AI citation. The goal was to become the company credible sources describe.

5:30 Measure honestly

How do you measure all this without fooling yourself? Three instruments. First, Search Console. AI feature clicks are part of normal web search totals, and the Generative AI performance report, rolled out to all sites by the end of August twenty twenty-six, shows impressions for AI Overviews and AI Mode by page, country and device. Second, a GA4 channel group for AI assistants, using a regex on session source. And third, a fixed prompt panel of real buyer questions, checked monthly in two or three assistants. Treat that panel as qualitative. Answers vary by user, place and time.

6:13 Watch me do it: AI visibility baseline

Watch me do it. I'm going to set up a prompt panel and a GA4 channel for AI assistants for a real business, a Manchester accountancy firm. First, I open a new sheet and write twenty questions buyers actually ask, taken from the firm's enquiry emails. Things like do I need an accountant for a limited company, and how much does a small business accountant cost in Manchester. Next, I add columns: date, assistant, answer summary, sources cited, firm mentioned yes or no, facts about us correct yes or no. Now I run the first five questions in two assistants and paste in what I see. The firm is mentioned once, via a local business magazine article. Good to know. Then I switch to GA4. Admin, data display, channel groups. I copy the default group and add a channel called AI assistants, with session source matching the regex from the lesson, and I drag it above referral. Save. Finally I open Search Console and check the Generative AI performance report for the firm's guide pages, noting impressions by page. I write the date on everything. Next month I repeat exactly the same questions, and that consistency is what turns noisy answers into a trend I can actually report.

7:43 Common mistakes

Common mistakes. Buying AI citation packages. Assuming unlinked mentions are worthless because there's no hyperlink. Treating one assistant's answer on one day as a trend. And creating llms dot txt files expecting Google rankings to move. Google Search ignores them. One more subtle mistake. Letting facts about your business go stale. Answer engines repeat what the web says. If your prices, hours or policies are out of date on your own site or major profiles, that's what gets summarised.

8:17 Recap

Recap. Google's AI features run on the same foundations as Search, so earned links and reputable mentions help in both places. Query fan-out means deep pages can be cited for sub-questions. Inauthentic mentions and attempts to manipulate AI answers fall under spam policies. And measurement combines Search Console, a GA4 channel group and a monthly prompt panel. Try this now. Write down ten real questions your buyers ask, run them in two assistants, and note which sources are cited. That list is your next PR target list. For the deep dive, take the AI Search Optimization course next.