Source: raw/How_to_Get_AI_to_Recommend_Your_Brand_in_16_minutes_AEO_Playbook.md (https://www.youtube.com/watch?v=L-X6HIrzrBI) + raw/AI_Search_-_Make_Your_Brand_the_1_Choice_Copy_These_SEO_Methods.md (https://www.youtube.com/watch?v=ZeMm6ZmiXeM) Creator: Grace Leung Platform: YouTube Duration: 16:35 (Aug 2026 AEO playbook) + 18:00 (Dec 2025 AI-search methods) Published: 2026-08-26 and 2025-12-05 Skill level: Intermediate (marketer comfortable with SEO tools; no code)

Grace Leung, an AI-for-marketing educator, gives two versions of the same method for getting answer engines (ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode) to cite and recommend a brand. Both split the job into the same parts: make AI understand the brand, make your content worth citing, and get other sites to vouch for you. The December 2025 video adds a technical “don’t get rejected” layer. Both videos are sponsored (the August one by HubSpot, whose AEO tracker she demos throughout; the December one by Semrush, demoing Semrush One), so the tool walkthroughs are vendor demos. This article is a practitioner playbook, not a study: the section below checks each tactic against the measured studies already in the wiki.

Key Takeaways

  • Three strategies (Aug 2026): (1) a clear, consistent brand identity, (2) own one topic with original depth, (3) engineer third-party mentions. Her one-line summary: AI “trusts what others say about your brand far more than what you say about yourself.”
  • Three “relationships” (Dec 2025): make AI love your content (so it uses you as a source), trust your brand (so it recommends you), and not hate your site (so it can crawl you). The first two map onto strategies 2 and 3 above; the third is a technical audit.
  • Brand identity is a reference check. Answer three things in one or two sentences each: who you are (category, what you do), who you are for (the ICP), and the two or three differentiators you want AI to repeat. Then make those answers match word for word on your site, social profiles, Google Business Profile and directories. Mismatched descriptions lower the engine’s confidence.
  • Audit what AI already thinks. Her client, an e-commerce home-improvement retailer, was described by AI as a manufacturer, so AI skipped it for “where to buy” questions. Ask each engine what your brand does and who it is for; check competitors too.
  • Rank position is the wrong AEO metric. She cites a SparkToro study (3,000 prompt runs across ChatGPT, Claude and Google AI) where the chance of the same brand list appearing twice was under 1 in 100, though one brand held 97% visibility. Her conclusion: measure how consistently you are considered, not where you rank. (Study figures unverified in the wiki; see Open Questions.)
  • Own a niche, not a volume target. Her definition: the set of prompts your buyers ask from research to decision. Map questions by buying stage (first hitting the problem, comparing options, right before choosing), pick the lane you can go deeper on than anyone, and give every piece something that exists nowhere else: your own data, results or take.
  • Information gain is the bar. She cites Google’s information-gain patent: search can prioritise sources that add information the user hasn’t already seen. Her caveat is that this doesn’t mean every LLM uses Google’s scoring, but a page that repeats the top results “is basically zero.” That is why mass AI-generated content won’t work without focus and original depth.
  • Listicles get cited, but treat them as a short-term tactic. In her HubSpot citation data, “best of” and top-list pages were the most-cited content type, and she has had clients mentioned through them. She calls this “short-term, but not long-term”, and says any page built for citations must add information the cited pages lack.
  • Third parties carry trust. She cites an AirOps “State of AI Search” figure of 80-90% of AI references coming from third-party sources, plus Google’s Search Quality Rater Guidelines, which tell raters to be sceptical of what a site says about itself and to look for independent reviews. Her advice: find where AI pulls answers in your category (review sites, listicles, social posts, Reddit threads), contribute genuinely helpful content, and pitch the roundups.
  • Don’t spam Reddit. The HubSpot tool flags high-intent Reddit threads to amplify. Leave helpful answers instead of promoting directly, which “might risk being banned.”
  • Technical floor (Dec 2025): check that robots.txt doesn’t block AI bots (her demo site blocked Gemini), keep the site structure flat with crawl depth of three or less, and add the key schema markup.
  • Refresh what you have. Update outdated statistics, add a new perspective and refresh the publish date. She says AI rewards freshness.

What the Measured Studies Say About Her Tactics

Her tacticVerdictEvidence in the wiki
Rank in classical search first (“strong correlation between Google rankings and AI citations”, Dec 2025)Backed, with a caveatSearch Rank scores 9.4 of 10 in Zyppy’s meta-analysis; AirOps finds retrieval rank 1 cited 58.4% vs 14.2% at rank 10 (AirOps). But Ahrefs’ March 2026 update shows AI Overview overlap with the top 10 falling from ~76% to ~38%, which postdates her December video. Ranking for fan-out queries now carries more of the weight. ^[inferred]
Brand/entity consistency as the first fixPlausible framing, weak measured evidenceSimilarweb’s “understand + trust” model puts on-page positioning first (Similarweb GEO playbooks). Zyppy scores Entity Consistency only 5.8 and Brand/Entity Trust 6.8, mid-to-low in its 23 factors. Worth doing as a cheap first step, but it is not the strongest measured lever. ^[inferred]
“Better SEO” (traffic, authority score) didn’t win topic ownership in nearly half of topicsConsistentHer claim is about domain strength, not rank. Zyppy scores Domain Authority 5.0 (“often weak”) and AirOps finds no positive DA correlation in ChatGPT.
Own one topic in depthBackedTopic Cluster Ranking scores 8.9 (Zyppy). AirOps finds pages covering 26-50% of fan-out sub-queries beat pages covering 100%. Similarweb: “scattered websites stay invisible.”
Information gain / original dataBackedSimilarweb information gain: the median top-3 page carries only 4 unique data points, and high-information-gain pages overperform. NP Digital lists first-party research among the four content types that earn citations (NP Digital).
Click and engagement behaviour as the “ultimate ranking factor”Backed for Google; speculative for other enginesShepard on click signals documents the antitrust-trial and API-leak evidence. Her claim that OpenAI and Perplexity built browsers to collect behaviour data is her own inference, which she flags herself (“they didn’t say that”).
Listicles get citedBackedIntent-Format Match scores 9.0 (Zyppy): listicles win “best” queries. NP Digital’s comparison and alternatives pages make the same point.
Third-party mentions and reviewsBackedNP Digital: ~80% of what AI cites is not your site. Similarweb calls co-citation “the gold standard of AI trust.” Shepard describes recommendation as “distributed consensus” (Zyppy AI visibility).
Rankings are random, so track consistencyBacked directionallySISTRIX drift study: AI Mode rotates 56% of cited domains weekly and ChatGPT Search 74%, while a brand’s own domain stays anchored. The SparkToro numbers themselves are not in the wiki.
Don’t block AI bots in robots.txtBackedURL Accessibility is Zyppy’s top factor (9.5). Zyppy’s audit checklist names the crawlers to allow (Zyppy playbook).
Add schema markup for AI extractionEvidence mixed ^[ambiguous]Zyppy scores Structured Data 5.6 (small but consistent correlation). AirOps finds +6.5pp correlational. Ahrefs’ causal study finds no significant lift from adding JSON-LD to pages already being cited. Add schema for classical rich results; don’t expect it to move AI citations alone.
Flat structure, crawl depth of three or less, “because some AI engines rely on Google’s index”UnverifiedNo study in the wiki measures crawl depth against AI citations. Google’s own guide says AI Overviews and AI Mode draw on the Search index (Google’s guide), which supports indexability in general, not the depth-3 rule.
Refresh stats and dates for freshnessBackedFreshness scores 7.0 (Zyppy); AirOps finds 30-89-day-old pages cited 32.8% vs 27.6% for 5+ years.

Try It

  1. Run the brand audit this week. Ask ChatGPT, Gemini and Perplexity: “What does [brand] do and who is it for?” If the answers are vague, wrong or different from each other, fix that first.
  2. Write the three identity answers (category, ICP, two or three differentiators) and paste them word for word into your site’s About page, social bios, Google Business Profile and directory listings.
  3. Map buyer questions by stage. Use an AEO tracker’s suggested prompts, or ask an LLM to generate awareness, consideration and decision-stage prompts for your ICP. Review them and make them specific to your business.
  4. Pick one lane and publish something only you can say on it: your own data, client results or a stated position.
  5. Read the citation panel, not just the visibility score. Find which channels feed the answers (peer sites, earned media, UGC), which content types are cited (often listicles), and which URLs. Then pitch those exact roundups, offering unique data, updated statistics or an expert quote rather than just asking to be included.
  6. Do the technical pass from the December video: robots.txt AI-bot access, crawl depth, and key schema types. Treat the schema step as low priority given the evidence table above.
  7. Reverse-engineer a cited answer with NotebookLM (Dec 2025 workflow). Copy the source URLs from an AI response for a target prompt, import them into NotebookLM, and ask it to analyse them and write custom instructions for an AI-search-optimised piece. For a prompt where a competitor outranks you, upload your own page too and ask what’s missing. Use the Studio panel to draft. Then add your own experience and perspective; she warns against publishing the draft as-is.

Implementation

Tool/Service: HubSpot AEO (Aug 2026); Semrush One = SEO Toolkit + AI Visibility Toolkit (Dec 2025); Perplexity Comet or another agentic browser; Gemini; NotebookLM. Setup:

  • HubSpot AEO: enter your website, accept or edit the suggested competitors, then define persona (ICP) + product combinations, because answers vary by who is asking (her example: a UK freelancer and a US enterprise CMO get different recommendations). The tool proposes prompts tagged by buying stage, scores visibility per engine (below 21% shows red; her own brand sat at ~15%), shows citations by channel and content type, and turns findings into recommendations typed as net-new content, social amplification or outreach.
  • Semrush One (Dec 2025 demo): Domain Overview now includes an AI visibility score. Keyword Gap against one close competitor, filtered by topic keyword or by page-2-to-3 positions, gives a target list. Export it to Perplexity Comet and ask it to review each SERP and produce a per-keyword optimisation checklist. For a new site, Keyword Strategy Builder clusters around a seed keyword (her demo analysed ~2,000 keywords) and Gemini prioritises the clusters against a brand overview document. For question research, use the Keyword Magic Tool’s Questions tab or Prompt Research (a database she describes as 100M+ prompts, filterable by ChatGPT, AI Mode and AI Overviews). Competitor Research and Brand Performance show share of voice, sentiment and “narrative drivers” per engine. Top Citations, filtered to Reddit, YouTube, LinkedIn and Quora, gives outreach and participation targets. Site Audit has an AI-search health panel covering bot blocking, crawl depth and markup. Cost: HubSpot AEO has a 28-day free trial with no other HubSpot subscription needed; Semrush offered a 14-day trial through her link. Paid pricing is not stated in either video. Integration notes: She warns against accepting the tool’s action list blindly. The recommendations are only as good as the brand, ICP and prompt setup you fed it.

Open Questions

  • Unverified statistics she quotes. None of these appear in the wiki as primary studies: Semrush “LLM visitors convert 4.4x better than organic”; the SparkToro 3,000-run study (<1 in 100 repeat chance, one brand at 97%); the Semrush + Kevin Indig study of 50,000 brands across 1,000+ topics over six months (“nearly half” won by the brand without the better SEO); AirOps “State of AI Search” 80-90% third-party share. NP Digital’s ~80% off-site figure is the closest corroboration for the last one.
  • Client claims are anecdotal. “Conversions from answer engines three to four times organic” and “already a top three revenue channel” come from one unnamed client. Compare Datos + SparkToro, which puts AI tools under 2% of desktop visits. A high conversion rate on a small channel can be true at the same time.
  • The Quality Rater Guidelines claim (raters told to search for reviews of the site to judge reputation) is stated from memory in the video and is not checked against the guidelines in the wiki.
  • Demo site name. The December demo optimises an AI website builder that the captions render as “type dream.com”; the real name is unclear.
  • Sponsor influence. Both tool walkthroughs are paid placements. Her rating of HubSpot AEO as having “the best UX so far” is not compared against a named alternative.