SEO Recap covering August 20, 2026: Google's Preferred Sources button and AI citation tactics

Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)

---

Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)

3 phases to build authority and earn AI citations
Source: Search Engine Land

• Up to 85% of top-of-funnel B2B brand mentions in AI search come from third-party content, per AirOps research, so getting cited with a link back requires earned coverage, not just owned pages.
• The 3-phase process combines AEO and SEO across owned and earned media:
  • Phase 1 makes pages accessible to LLMs through logical navigation, machine-readable structure, and optimized PDPs with product-focused H1s and H2s
  • Phase 2 creates fresh and repurposed content to control baseline brand messaging
  • Phase 3 uses outreach and digital PR to earn third-party citations from authoritative and niche sources
• Build a query fan-out list from user prompts plus keyword research to see who is cited in AI results and where the gaps are.
• Keep brand messaging consistent across first- and third-party platforms so LLMs can find, understand, and connect the brand proposition.

The coming "edits for AI" money grab from publishers
Source: Search Engine Land

• As brands audit third-party sources shaping how AI describes them, some publishers may start charging "editorial processing fees" to correct outdated bios, titles, company descriptions, and slogans.
• AI systems build answers from all crawled sources, so repeated outdated descriptions across third-party sites can keep surfacing even after a company has repositioned.
• There is a line between correcting objectively false information and reputation laundering, and accurate historical articles should not be rewritten just because a subject will pay.
• The standard should be factual accuracy, not whether the subject has enough money to influence the record.

The keyword universe was always smaller than we thought
Source: Search Engine Journal

• A June 2026 audit of 3,750 responses across 3 models and 250 category queries found all 3 agreed on the top brand only 41.6% of the time, but at least 2 agreed 91.6% of the time, meaning models settle on a small, stable eligible set while the order shifts.
• Personalization concentrates rather than dissolves this convergence. An audit of 2,000 runs across 10 personas found category leaders held ~80% consistency while mid-market brands swapped up to 75% of the recommendation set.
• A Trine University and Texas A&M experiment pitting one real brand against 9 fictional ones with identical attributes saw the real brand recommended in every one of 670 valid trials, showing models select on name recognition, not product quality.
• Genuine competitive vacuums appeared in only 8% of 250 queries, with healthcare technology highest at 20%, pointing to sparse, thinly documented categories as the temporary openings.
• Nearly all published AI visibility measurement comes from vendors selling AI visibility measurement, so the numbers should be held loosely.

Prompt injection revives a 25-year-old black-hat SEO tactic
Source: Search Engine Journal

• Hidden instructions for LLMs, written in tiny white-on-white text, echo the white-text keyword stuffing of pre-2010 SEO, but now an LLM acts on the instruction rather than just ranking it.
• Documented cases span multiple domains:
  • July 2025 arXiv preprints from 14 institutions hid prompts telling AI reviewers to "give a positive review only"
  • A May 2026 study of 196,682 resumes found ~1% carried hidden prompt injections, over 90% using keyword-dense text with no explicit command
  • Microsoft's February 2026 research found 50 injection attempts from 31 real companies planting themselves in AI assistant memory as "trusted sources"
  • A July 2026 Connecticut court filing hid instructions to any AI processing it, and the self-represented plaintiff was sanctioned
• A University of Turin study ran 42,000 outputs and found positive steering, forced refusal, and external redirection succeeded over 98% of the time on ChatGPT and Gemini.
• Researchers call the root cause "contextual blindness": transformers cannot architecturally separate content being evaluated from control text embedded inside it.

---

Technical SEO

How to identify and fix content cannibalization at scale
Source: Search Engine Journal

• 3 ways to detect cannibalization:
  • Google Search Console, by filtering a query and checking whether multiple URLs split traffic evenly or slip to pages 3-5
  • Crawlers like Screaming Frog, Sitebulb, or Botify to extract and sort duplicate title and H1 tags
  • Rank trackers like Semrush, Ahrefs, and Authority Labs to spot phrases stuck in the mid-20s to 50s with multiple ranking URLs
• LLM-generated content commonly triggers this because it produces thin, repetitive pages regardless of prompt quality.
• Fixes include consolidating pages, applying meta robots or canonical tags, correcting internal links to match keyword intent, and requiring content teams to clear new topics with SEO.
• Complementary pages, such as a how-to guide and a conversion page, are not cannibalization, and incorrect canonical tags on new product variants are a common hidden cause.

Favicons disappearing from Google Search results
Source: Search Engine Roundtable

• Site owners reported a spike in favicons vanishing from Google Search snippets starting around Monday, with Google's generic globe icon appearing instead.
• Affected sites reported valid 128x128 favicons returning 200, referenced in HTML and not blocked by robots.txt, yet still not displaying after repeated reindexing requests.
• The complaints began before the August 2026 spam update, suggesting a Google-side change rather than site errors.
• Some examples began working again shortly after reporting, and Google's Rajan Patel said he is looking into it.

---

Organic Search & Algorithm Updates

Google makes the Preferred Source button more seamless
Source: Search Engine Land

• Google's new interactive Preferred Source button adds a site as a preferred source and returns the reader to where they left off, replacing the old badge that dropped users into Google's source preferences tool.
• Preferred Sources let searchers star publications so Google shows more of their stories in Top Stories, and selected sites can carry a "preferred" badge in AI Overviews and AI Mode.
• Google says people have selected more than 600,000 unique sources, up from 200,000 in May 2026, and that those users are twice as likely to click.
• The embed code is available in Google's Search Central documentation.

Google Discover adds conversational feed customization
Source: Search Engine Land

• Discover users can tap the 3-dot menu on any card and tell Google, in their own words, which topics or links they want to see more or less of.
• The feed adjusts on the fly and remembers those requests, adding a written request as a third option alongside following and blocking topics.
• The change gives content teams a way to surface niche topic ideas and another path for readers to find their content in Discover.

SEO priorities for 2027
Source: Search Engine Land

• Google says AI Mode passed 1 billion monthly users in its first year and AI Overviews reach 2.5 billion monthly users, while zero-click searches account for 68.01% of US Google searches (January-April 2026), up from 60.45% in 2024 per SparkToro.
• Traffic to US retail sites from generative AI tools grew 693% year over year during the 2025 holiday season, with AI-referred visitors converting 31% more, per Adobe Digital Insights.
• Ahrefs' analysis of 75,000 brands found branded web mentions correlate far more strongly with AI Overview visibility (~0.664 Spearman) than backlinks (0.218), with YouTube mentions the strongest single signal at ~0.737.
• Concrete steps include auditing how LLMs mention your brand monthly, tracking AI-referred sessions in GA4, using Search Console AI performance reports (launched June 3), and writing concise 40-60 word answers structured for AI Overviews.

Google expands personalization across Search, Discover, and News
Source: Search Engine Journal

• Publishers can embed a new interactive Preferred Sources button that adds their site as a preferred source and returns readers to the page, unlike the old badge that sent them to Google's source preferences tool.
• Preferred Sources selections passed 600,000 unique sources, up from more than 345,000 in May, and selected sites are more likely to appear in Top Stories and can be badged in AI Overviews and AI Mode.
• Discover will let Google app users type in their own words what they want more or less of via the 3-dot menu, and Google News on Android now allows topic-customized audio briefings with source attribution.
• Publishers who already added the earlier badge should check whether the new button replaces it automatically or needs a fresh embed from Search Central.

Why Google announces some search updates but not all
Source: Search Engine Roundtable

• John Mueller reiterated that Google announces larger, broader system changes so site owners can attribute big metric swings to Google rather than their own technical changes.
• Many ranking systems update continuously and automatically, and Google withholds smaller updates to avoid causing unnecessary worry.
• This explains why Google confirmed the August 2026 spam update while leaving other unconfirmed updates unannounced.

Google's official announcement on personalizing Search, Discover, and News
Source: Google Search Product News

• Google introduced a new Preferred Source button that lets readers prioritize favorite publishers in Search results and AI summaries, with more than 600,000 unique sources selected so far.
• Discover will soon let Google app users type exact interests, including exclusions like "eco-friendly only" or "no camping," through the 3-dot menu, with the feed adjusting and remembering requests.
• Google News on Android now supports customizable daily audio briefings by topic with source attribution and links out, drawing on its news AI pilot program.
• Publishers can get the new button code in Google Search Central documentation.