There is no shortage of AI search advice right now. Frameworks, acronyms, case studies claiming 800% citation increases. Most of it does not get stress-tested. It just gets shared.
Harpreet Singh does it differently. He runs actual experiments to check whether the claims flying around LinkedIn and X hold up. In this SEOTalk webinar, he sat down with Parth & Malhar to cut through the noise and talk about what is actually working, what is a distraction, and where brands are quietly setting themselves up for problems down the line.
Here is what came out of that conversation.
The Acronym Problem Is Not Really the Problem
GEO, AEO, LLMO, AI SEO. The naming debate has been running for a while now and Harpreet’s position is clear: the names do not matter. The tactics underneath them do.
His framing is useful. Think back to when TikTok SEO and YouTube SEO became terms. Nobody created new acronyms for those. The work still mapped back to understanding search behavior, creating content that satisfied it, and building authority around it. AI search is no different. In his estimate, 70 to 80 percent of what goes into AI search optimization is just SEO, done well and kept current.
The reason new terms keep appearing is partly commercial and partly cultural. SEO has a perception problem with some executive teams. It does not excite the C-suite the way “AI visibility strategy” does. Marketing professionals have picked that up and run with it. The framing changes. The fundamentals do not.
That said, something has changed about scope. SEOs are now in conversations they were not part of two years ago: social media planning, video strategy, product team discussions. The AI search brief has expanded the seat at the table. That part is new, even if the core work is not.
What to Actually Do Right Now: The Short List
Harpreet is selective about short-term tactics. His view is that most of what gets promoted as AI search tactics is either redundant with good SEO practice or actively risky. Here is what he recommends instead.
Invest in video. This is his clearest, most repeated recommendation. YouTube is the highest-cited domain in Google AI Overviews across most query categories. If a brand is not producing video, it is absent from the surface that matters most in AI-generated answers right now. The barrier is lower than most teams assume. An iPhone and a 10-minute recorded script is a starting point. The strategy matters more than the production budget.
One nuance worth noting: most of the YouTube videos being cited in AI Overviews are not from brand-owned channels. They are from third-party creators with topical authority in the category. That means creator partnerships and influencer collaborations are not peripheral brand activities anymore. They are a direct input to AI citation strategy.
Publish video everywhere. Once a video exists, distribute it across YouTube, Instagram, X, TikTok, and wherever the audience is. Different AI systems draw from different sources. Broader distribution increases the surface area for being picked up.
Work third-party domains strategically. Look at what is ranking for your key queries in AI surfaces. Then figure out if you can create content on those platforms. LinkedIn articles, relevant community contributions, and guest content on high-authority sites all feed into how AI systems source answers.
Focus on Google first. AI search strategy should be optimized for Google’s AI Overviews and AI Mode. Google will remain the dominant search platform by a significant multiple. More importantly, tactics designed for ChatGPT or Perplexity can sometimes create problems with Google’s algorithm. Getting the Google strategy right tends to take care of the other channels. The reverse is not always true.
What to Stop Doing
Harpreet was direct on this: the bigger risk for most teams right now is not missing new tactics. It is wasting time and budget on things that either do not work or create future problems.
Chunking is not a thing for AI search optimization. It is a concept from model training, not a content publishing tactic. Readable paragraphs, clear headers, and logical structure have been best practice since before most current practitioners started in the industry. If a tool vendor is selling you “content chunking” as an AI search feature, that is worth interrogating.
Scaling content with AI at volume creates long-term risk. It might spike citations in ChatGPT in the short term. It might even temporarily increase traffic from Google. But the pattern Harpreet is observing follows a predictable arc: peak, then drop, then a recovery that takes longer than the gain was worth. If a business is currently growing 15 to 20 percent year on year, publishing two thousand AI-generated posts is a gamble on the downside, not an investment in the upside.
LLMs.txt, FAQ schema stacking, and similar tactics are generally not worth team time. The real cost is not just the direct effort. It is the meeting time, the internal alignment conversations, the distraction from higher-value work. Five people spending an hour debating whether to implement an LLMs.txt file represents a significant business cost before a single decision is made.
Self-promotional listicles are becoming a liability. Google has been actively working to reduce the impact of best-of listicles in AI Overviews. Harpreet cited a case study from a finance startup that built its AI citation strategy almost entirely on listicles. The pages still get cited in AI Overviews, but they are not mentioned in the actual AI-generated answer. The company is reinforcing competitors’ positions while receiving no meaningful benefit. That is a net negative result dressed up as a visibility win.
The Press Release Experiment and What It Reveals
One of the most instructive moments in the webinar was Harpreet walking through a test he ran to demonstrate how easily AI search systems can be influenced.
He created a fictional company in Vancouver, wrote a press release calling it the best AI agency in the city with a five-star rating, and distributed it through standard press release channels. Within a short period, the claim was appearing in Google AI Overviews and ChatGPT responses, unsupported by any verified reviews or third-party validation. The syndication across hundreds of sites created enough signal for the AI systems to treat it as fact.
He ran a similar experiment with a newsletter, positioning it as the best AI SEO newsletter for CMOs. The AI overview recommendation persisted long after the test was complete, despite no credible basis for the claim.
His point is not primarily about tactics. It is about the nature of these systems. AI search does not verify. It aggregates. The more sources repeat a claim, the more weight it carries. That creates a genuine manipulation opportunity that is currently being exploited, particularly in press release distribution. It also means users should be more skeptical consumers of AI-generated answers than most are.
For practitioners, it opens a legitimate strategy question around digital PR. Harpreet’s suggestion: look at how established PR companies are constructing press releases right now. The keyword placement, the strategic positioning, the specific claims being made. It is instructive.
How to Handle the Small Business Question
A question from the webinar audience asked how small businesses with limited budgets should approach AI search relevance. Harpreet’s answer was practical and does not require external spend to start.
Go to ChatGPT. First, search your business name with web search disabled and ask what the model knows about you from its training data alone. That is your baseline. Then open another tab, enable web search, and ask the same question. Compare the two answers.
What you are looking for: does the AI describe your business accurately? Is it pulling the right information or something from a competitor’s site or an outdated news story? Are there negative reviews surfacing, and if so, from which platforms?
The gap between those two answers tells you where to focus. If the messaging is wrong, publish content on your own site that clearly addresses what the business does. If bad reviews are surfacing from specific platforms, that is where your review management effort goes. If a competitor’s content is being pulled in to describe your business, that is a content gap to close.
The principle: start with how AI currently perceives your brand. Fix that before worrying about non-branded visibility. Getting the brand picture right is the foundation everything else builds on.
The Long Game Most Brands Are Getting Wrong
Harpreet’s clearest critique is aimed at the time horizon problem. Most AI search optimization is being done with a six to twelve month view. The decisions that cause the most damage are the ones made without thinking about what they look like in year three.
The Shopify content strategy controversy was a reference point. When AI search tactics become visible enough that mainstream publications are writing critical pieces about them, the brand perception problem is now bigger than the SEO problem. Being perceived as a company that is spamming the internet in the name of AI visibility is harder to recover from than a traffic drop.
His recommendation for practitioners in these conversations: before you recommend a tactic to a client or internal stakeholder, think about the long-term impact on the business. Not just the SEO impact. The business impact. If scaling AI content will give a traffic spike followed by an algorithm penalty and a two-year recovery period, is that worth it for the incremental customer acquisition during the peak? Sometimes the answer is genuinely yes. But the framing of the question matters. Present the trade-off honestly, including the downside scenario. That honesty builds more durable trust with clients and stakeholders than any short-term result will.
Visibility vs. Traffic vs. Quality: The Actual Trade-off
Parth asked the core question directly: what is the biggest trade-off for a brand that over-optimizes for AI search citations at the expense of everything else?
Harpreet’s answer was precise. AI search currently accounts for one to five percent of traffic and revenue for most businesses. Organic search from Google still represents 30 to 60 percent. A brand that damages its organic SEO standing in pursuit of ChatGPT citation gains is trading a large, proven channel for a small, developing one. That is almost never the right calculation.
The citation metric itself is largely a vanity measure. One explainer page covering a basic definition can accumulate hundreds of citations. A product page that drives direct commercial consideration will have far fewer citations and far more business impact. Measuring AI citations without connecting them to the funnel position of the pages being cited is not a useful strategy signal.
The practical alternative: focus on your product pages and commercial intent content first. Make sure your G2 profile, Trustpilot listing, and other review platforms are accurate and well-maintained. Ensure that your brand description on third-party platforms matches what your own site says. These are the surfaces AI systems use when someone is actually evaluating a solution, not just asking a general question.
Mentions matter more than citations. If ChatGPT mentions your brand by name, a user evaluating options may search for you directly in their next query. An inline citation that links to your homepage from a brand mention is more commercially useful than a citation in a listicle where three competitor names appear above yours.
Budget Reallocation Worth Considering
One specific suggestion from Harpreet is worth pulling out separately. Many businesses are still spending on link building through direct acquisition. Before renewing that budget line, ask what it is actually delivering for both organic and AI search strategy.
An alternative use: take the same budget and commission an original data study on a topic relevant to the business. A study with proprietary findings gives you a distributable asset, a natural reason for journalists to reference the business, earned backlinks from editorial coverage, and genuine topical authority signals that feed both traditional and AI search. It is the kind of content that gets cited because it is the primary source, not because it was designed to be cited.
The Toronto road fatality database example he raised is instructive. Public data existed but no one had aggregated and presented it usefully. One piece of content filling that gap created a media-distributable asset with real link and citation value, at a cost comparable to a few months of link purchases.
The Practical Checklist
Pulling the webinar together into what to do and what to stop:
Start with how AI currently describes your brand, without web search enabled, and with it. Fix messaging gaps before optimizing for anything else. Invest in video now. Distribute it widely. Build creator relationships in your category. Prioritize Google AI Overviews as your primary AI search target. Use digital PR strategically for earned mentions in quality editorial sources. Maintain review profiles and ensure brand descriptions are consistent across every platform. Focus citation strategy on product and commercial pages, not top-of-funnel explainers.
Stop scaling AI content at volume without a clear quality bar. Stop treating citation counts as a meaningful KPI without connecting them to funnel position. Stop allocating meeting time to tactics like LLMs.txt and content chunking that have no demonstrated return. Stop building listicle strategies around your own product where competitors end up getting more benefit than you do.
The underlying principle throughout: think on a ten-year timeline for the business, not a six-month timeline for the campaign.
Harpreet Singh is an AI search consultant based in Vancouver with over a decade of experience across agency and in-house roles. He publishes the SEO Espresso newsletter and shares test-driven findings on X at @harpreetchatha_ and on LinkedIn. This session was part of the SEOTalk Webinar Series, hosted by Parth Suba and Malhar Barai.