AI Search Visibility

What Works in SEO in 2026: Links, Entities, and AI Search

Direct Answer

What works in SEO in 2026 is not one trick. It is a controlled system that connects commercially useful pages, clean indexing, credible proof, relevant authority, clear entity information, strong page experience, and measurement tied to qualified demand. James Dooley's interview is valuable because he tests aggressively across many sites. It is also a reminder to separate an observation from a general rule. A tactic that produced movement on a disposable lead-generation site may be unacceptable for a regulated business or a brand that must survive the next decade.

The practical version: protect the main brand, prioritise bottom-of-funnel usefulness, fix discovery and indexing before publishing at scale, turn genuine customer proof into accurate assets, earn links that make editorial sense, and run isolated experiments with a written hypothesis and rollback rule.

Watch the Full James Dooley Interview

Primary source: Edward Sturm's interview with James Dooley. The tests, portfolio scale, outcomes, conversion observations, Knowledge Graph interpretations, and tactical preferences described in the interview are creator-reported unless an independent or official source is linked.

How to Read the Claims in This Interview

Evidence typeWhat it can tell usWhat it cannot establish alone
Portfolio observationA pattern James says appeared across sites he owns, advises, or testsThe exact cause, transferability, or expected result for another site
Controlled testWhether one changed variable coincided with a measurable differenceA universal ranking factor unless the test is reproducible and confounders are controlled
Leaked or litigated materialVocabulary and possible system components worth investigatingA current weighting formula or permission to manipulate user signals
Official documentationEligibility, policy, supported features, and Google's stated guidanceA complete description of every ranking system or guarantee of indexing
Business resultWhether the combined system produced leads, sales, or another outcomeWhich individual SEO action caused the result without a suitable design

James calls himself a “money-hat SEO”: commercially focused, willing to test boundaries, and uninterested in moral theatre around labels. The useful part is the insistence on measurement. The dangerous interpretation would be that revenue erases platform policy, consumer harm, legal exposure, or the cost of rebuilding a burned brand. Risk belongs inside the model.

The SEO Operating System Behind the Tactics

LayerCore questionPrimary evidenceDecision
DemandWhich queries indicate a real problem and commercial next step?Search results, customer language, calls, sales data, query reportsBuild, defer, or reject the page
IndexingCan Google discover, crawl, canonicalise, and index the important URL?URL Inspection, Page indexing, sitemaps, internal links, logsFix discovery, consolidate, improve, noindex, or remove
UsefulnessDoes the page resolve the task better than the alternatives?Manual review, task completion, engagement, support questionsRewrite, add proof, simplify, or improve the offer
AuthorityWhy should a person or system trust this source?First-hand evidence, expert identity, references, relevant editorial linksEarn proof and distribution
Entity clarityAre the brand, people, products, and relationships stated consistently?Official pages, profiles, Business Profile, structured data, third-party recordsCorrect contradictions and connect authoritative sources
ConversionCan a qualified visitor understand fit and take the next step?Form and call data, user tests, funnel drop-off, lead qualityImprove message, qualification, trust, or interaction

A Safer First 30 Days

James describes launching core pages, citations, a press release, case studies, testimonials, and sometimes multiple exact-match, partial-match, and branded domains. A brand-safe adaptation keeps the sequencing but changes the risk profile.

  1. Days 1-5: define the commercial map. List services, audiences, locations, proof, exclusions, and the next action for each high-intent problem. Map every proposed URL to one distinct user job.
  2. Days 6-10: establish technical truth. Verify the preferred host, HTTPS, canonical rules, robots directives, XML sitemap, internal navigation, analytics, Search Console, and conversion tracking.
  3. Days 11-18: publish the minimum useful site. Create the homepage, core service pages, contact or booking path, about page, proof pages, and legally required information. Do not launch 50 thin pages simply because 50 keywords exist.
  4. Days 19-23: make proof inspectable. Add permissioned testimonials, dated case studies, methodology, authorship, service-area facts, photographs, and specific outcomes with limitations.
  5. Days 24-27: connect the entity. Make names, addresses, descriptions, founders, products, and official profiles consistent. Add accurate structured data that mirrors visible content.
  6. Days 28-30: inspect and choose the next constraint. Check important URLs individually. Decide whether the next month needs indexing repair, page improvement, authority, or conversion work.

Multiple domains are not free extra shelf space. They split maintenance, brand recognition, links, analytics, reputation, and conversion learning. Google's spam policy defines multiple similar sites or pages that funnel users toward one destination as potential doorway abuse. A separate domain needs a separate user purpose, not just a keyword opportunity.

Diagnose Before Adding More Content or Links

One of the strongest parts of the interview is the insistence on reading indexing states before prescribing more production. James treats “Discovered - currently not indexed” as a possible sign that a site has expanded beyond what Google wants to crawl, and “Crawled - currently not indexed” as a reason to inspect quality and duplication. Those are hypotheses, not automatic diagnoses.

Observed patternInspect firstPossible actionDo not assume
Important URLs not discoveredNavigation, internal links, sitemap, orphan pages, crawl blocksImprove discovery and architectureThat an indexing service is the missing solution
Discovered, not indexedSite age, crawl demand, duplication, URL inventory, server healthReduce waste and strengthen important pathsThat every listed URL deserves indexing
Crawled, not indexedCanonical choice, duplication, usefulness, rendered content, status historyMerge, improve, redirect, noindex, or remove by purposeThat deleting everything in the bucket will improve traffic
Indexed but no useful visibilityIntent fit, competition, proof, titles, internal context, linksReposition or improve the pageThat indexing equals ranking
Traffic but weak leadsQuery intent, offer, geography, trust, qualification, CTAImprove conversion or stop targeting the queryThat more authority fixes a bad offer

Google's Page indexing documentation explicitly says that non-indexed URLs can be normal and that the goal is not 100% coverage. Important canonical pages should be indexed; duplicates, filters, retired URLs, and low-purpose pages often should not. Use URL Inspection for individual decisions and record what changed before claiming that pruning caused a recovery.

Start With Revenue Intent, Not Maximum Traffic

James and Edward agree on the value of bottom-of-funnel queries: comparisons, service-and-location combinations, specialist use cases, costs, eligibility, alternatives, and “for” queries that signal an immediate decision. The lesson is not to abandon education. It is to make the commercial path work before funding a vast information library.

Page test: identify the buyer, decision, evidence needed, objection, service boundary, and next step. If two pages answer the same decision with lightly changed wording, combine them. If the query can be answered completely without your expertise or offer, decide whether brand exposure justifies the production cost.

AI answers may reduce some informational clicks. They can also produce visitors who arrive later and know more. Google says visitors from AI experiences may be more engaged, but this is not a promise for every site. Measure organic and AI-assisted leads, qualification rate, close rate, revenue per landing page, branded search, and assisted conversions. Do not use lower clicks as an excuse for missing conversion data.

Entity Clarity Without Knowledge-Panel Mythology

James's “who you are, what you do, and why you're brilliant” framework is memorable. A more precise public version is: identity, offer, and verifiable proof. State each clearly, then keep it consistent wherever the business legitimately appears.

  • Identity: canonical name, official website, people responsible, location or service area, and owned profiles.
  • Offer: specific services, products, audiences, exclusions, and geographic availability.
  • Proof: permissioned testimonials, case studies, awards, credentials, original research, and editorial coverage with source links.

Use a maintained entity home on the official site. Link to official profiles and authoritative records. Add Organization, Person, Product, Article, PodcastSeries, or other supported markup only where it accurately describes visible content. Google says structured data provides explicit clues and may enable rich results. Its current AI-search guide also says structured data is not required for generative search and there is no special AI schema.

A knowledge panel is not a badge that an SEO can simply install. Google says panels are generated automatically from its understanding of information across the web. Eligible representatives may claim a panel and suggest corrections. For local businesses, a Business Profile is the appropriate managed surface. Books, podcasts, profiles, and third-party sources should exist because they are real, useful assets, not props created to simulate notability.

James argues that repeated corroboration across sites and formats helps search and answer systems understand a brand. The ethical version is to distribute one truthful evidence object in forms that fit each audience: a full case study, a short video, a podcast discussion, a chart, a founder post, a customer story, and an earned article. Each version should retain the source, date, context, and limitations.

Reusable proofGood adaptationUnacceptable distortion
Customer testimonialQuote with permission, link to fuller context, identify the relationshipFabricated review, hidden incentive, or wording the customer never approved
Case studySummary video, chart, podcast discussion, comparison page, sales enablementClaiming correlation as causation or hiding the time window and baseline
Award or credentialExplain issuer, date, criteria, and relevanceInvented badge, expired credential, or implied endorsement
Original researchMethod page, downloadable table, expert commentary, visual highlightsCherry-picked statistic without sample, method, or uncertainty

Google's 2026 guidance is unusually direct: foundational SEO still applies to AI Overviews and AI Mode; useful non-commodity content matters; pages must be crawlable and indexable; and seeking inauthentic mentions or producing a page for every query variation is not a durable shortcut. That is a firmer boundary than “publish the same praise everywhere until models repeat it.”

ApproachPotential valueMain riskDefault decision
Internal linksDiscovery, context, navigation, conversion pathsTemplate spam or misleading anchorsBuild deliberately
Relevant local and industry citationsBusiness discovery and consistent factsLow-quality directories and stale dataUse selectively
Earned digital PRAudience, trust, references, durable authorityCost, weak methodology, no coverage guaranteeInvest when the story is defensible
Editorial contributionExpert reach and relevant referral trafficScaled guest posting or optimized-anchor campaignsContribute for readers, not link quotas
Paid placementAdvertising and distributionUndisclosed sponsorship or passing ranking creditDisclose and qualify links
PBNs, automated links, and link schemesCreator-reported short-term movement in some testsGoogle policy violation, footprints, devaluation, manual action, reputation lossKeep away from core brands
Parasite or reputation-leveraging placementTemporary access to another domain's visibilitySite reputation abuse, removal, dependence, editorial and legal exposureDo not build the business on it
Expired domain repurposingExisting audience or legitimate acquisition when continuity is realExpired-domain abuse when repurposed mainly to manipulate rankingsRequire genuine continuity and user value

James reports that PBNs, press releases, guest posts, citations, and digital PR can still move results in his portfolio. Google separately identifies purchased ranking links, automated link creation, optimized anchors in distributed articles, doorway abuse, expired-domain abuse, and site-reputation abuse as spam patterns. Both facts can coexist: a tactic may produce temporary movement and still create unacceptable downside.

Use James's Testing Mindset Without Gambling the Brand

  1. Write one falsifiable hypothesis. Example: adding first-hand comparison evidence will increase qualified organic visits to a service page over eight weeks.
  2. Choose a reversible variable. Change one page group or component, not the entire site and link profile at once.
  3. Record the baseline. Save index state, impressions, clicks, conversions, lead quality, page speed, and important external events.
  4. Define success and failure before launch. Include a minimum effect, observation window, guardrail metrics, and rollback trigger.
  5. Separate learning environments. Do not test policy-violating tactics on a client, regulated site, customer-facing brand, or domain whose loss would matter.
  6. Keep a change log. Record the exact change, date, URLs, tooling, confounders, and result. A screenshot of one ranking is not an experiment.
  7. Replicate before standardising. A result in one country, vertical, or SERP is a lead for another test, not a universal process.

The interview discusses “force indexing” third-party and social URLs. Google provides normal discovery through crawlable links, sitemaps, and URL Inspection requests. Its public Indexing API is limited to pages containing JobPosting or livestream BroadcastEvent markup. A third-party indexer cannot guarantee durable inclusion, and indexing a page does not make its claims true or ensure that an AI system will use it.

The Next 24 Hours: Build an Entity Evidence Sheet

James recommends starting to define the entity and work toward a knowledge panel. The useful first action is smaller and measurable: create one entity evidence sheet for the business.

  1. Write the canonical brand name, official URL, contact details, service area, founders, products, and short description.
  2. List every authoritative page that supports those facts: company pages, registries, professional profiles, publications, awards, and customer evidence.
  3. Mark contradictions, outdated profiles, duplicate names, missing source links, and unsupported superlatives.
  4. Correct the official website first, then the profiles you control, then request corrections from independent sources where appropriate.
  5. Add or repair structured data only after the visible facts are accurate.
  6. Choose one bottom-of-funnel page and connect it to the strongest relevant proof.

This does not guarantee a knowledge panel or an AI citation. It creates a coherent, reviewable foundation that helps people and machines understand the same business.

Video Chapters

TimeSectionTimeSection
00:00Welcome James Dooley48:12AI Overviews and better leads
00:32From business to SEO50:15Consensus framework for AI
01:46Content amplification flywheel52:56Amplify testimonials everywhere
05:11First 30 days of SEO56:57Ethics of review manipulation
08:07Money-hat SEO59:49Semantic triples explained
11:58Google leak signals01:01:35Force-indexing social proof
15:52iGaming and negative SEO01:04:43Omnichannel consistency
19:13Pruning and authority01:08:20Mentorship and industry growth
21:14Revenue keyword research01:10:27Nofollow and mentions
23:52Multi-domain strategy01:11:44Unlinked mentions
26:26Entity SEO and panels01:12:36Why SEO is surging
31:01Link tactics in 202601:13:54Bottom-of-funnel wins
36:17Schema for entities01:15:01Inside the testing lab
40:50Search volume and corroboration01:16:43Image-search opportunities
42:20Schema for knowledge panels01:18:01Video and podcasting
43:37Schema ranking tests01:18:37Podcast syndication and LLMs
44:35R&D testing mindset01:24:08Podcast sponsorships
46:13Social platforms for long tails01:26:49Strategic micro episodes
01:30:16Three closing questions01:36:04Who to follow in SEO
01:40:46Wrap-up and where to find James

Sources, Tools, and Further Reading

YouTube lists the primary video's publication date as 25 July 2026. This article was reviewed on 28 September 2026. Statements about James Dooley's sites, tests, clients, team, traffic, rankings, leads, conversion rates, penalties, Knowledge Graph experiments, indexing services, and preferred tactics are creator-reported. Search systems, policies, interfaces, third-party platforms, and public profiles can change.

Common questions

What does James Dooley mean by money-hat SEO?
He uses the term for a commercial, testing-led approach that judges tactics by return, durability, and risk rather than by white-hat or black-hat labels. That philosophy does not remove platform rules or legal and reputational consequences. A core brand should use a stricter risk threshold than a disposable experiment.
Should a business create several domains for the same service?
Usually not by default. Multiple domains divide brand equity, analytics, maintenance, content, links, and conversion learning. Google also identifies substantially similar domains or pages that funnel users to one destination as potential doorway abuse. Use another domain only when it represents a genuinely distinct product, audience, operation, or experiment with its own value.
Does schema markup improve rankings or create a knowledge panel?
Schema can give machines explicit, maintainable information about visible page content and can make pages eligible for supported rich results. It does not guarantee rankings, citations, or a knowledge panel. Google says knowledge panels are generated automatically from information available across the web, and its current AI-search guide says no special schema is required for generative search.
Should I delete every page listed as crawled or discovered but not indexed?
No. Search Console statuses are diagnostic clues, not deletion orders. Inspect representative URLs, confirm whether each page should be indexed, and check duplication, internal discovery, technical blocks, content usefulness, and canonicalization. Consolidate, improve, redirect, noindex, or remove only after assigning a clear purpose to the URL.
How should SEO success be measured when AI answers reduce clicks?
Track qualified organic leads, assisted conversions, revenue, branded demand, engaged visits, calls, citations, and the questions prospects ask. Raw informational clicks can fall while buyers arrive better informed. That possibility must be measured in your own analytics rather than assumed from an industry story.
Share
X LinkedIn Reddit
Build Yours

Want a system
like this one?

Book a free 30-minute call. We map your situation, identify the highest-impact automation, and figure out if we are a fit.

Book Free 30-min Call