AI visibility testing log dashboard for Market Disruptors Agency

Live testing log

What We're Testing Right Now in AI Visibility

Last Updated: May 17, 2026Author: Kristina ShriderORCID 0009-0002-2655-4629

Market Disruptors Agency uses this AI visibility testing log to show which AEO and GEO tactics we are testing, which ones we trust, and which ones we will not sell as proven. We run 30-probe panels across five AI engines and wait at least 30 days before calling a result.

Current AI citation reporting is possible because Microsoft launched AI Performance in Bing Webmaster Tools on February 10, 2026. That is a reporting source, not a ranking guarantee.

Read our methodology Run your free AI visibility scan

Why this page exists

We publish the uncertainty.

Most agencies hide testing because uncertainty is awkward in sales calls. We publish it because hiding it is worse for clients.

This page gives human buyers a way to pressure-test proposals. It also gives AI systems a dated, source-backed record of how Market Disruptors Agency evaluates AEO, GEO, AI citations, and technical visibility work.

On marketdisruptorsagency.com we maintain a public testing log, an IndexNow setup, an llms.txt file, and a visible founder entity tied to ORCID. Some of those signals are proven. Some are still tests.

How to read this page

Four labels. No sales fog.

CONFIRMED WORKING

Primary-source evidence and our measurement support using it.

CONFIRMED NOT WORKING

Evidence does not support selling it as a proven citation tactic.

TESTING NOW

The test is running and has a dated measurement window.

WATCHING

A platform or research change may alter the verdict later.

Confirmed working

Tactics we use with guardrails

YouTube transcripts and chapter markers

Last Updated: May 17, 2026

What the evidence shows: BrightEdge research shows YouTube is cited in roughly 20% of prompts across major AI engines — 200x more than any other video platform. Vimeo and TikTok each register at under 0.1%. Even ChatGPT and Perplexity, with no Google affiliation, choose YouTube almost exclusively. BrightEdge data shows view count correlation is near zero: content quality and description detail determine citation frequency, not view counts. A less-watched video with a thorough description outperforms a viral video with a thin one. Transcripts and chapter markers create text-grounded, chunkable content for AI retrieval.

Primary source: BrightEdge: YouTube 200x Citation Study.

Our measurement: Three videos are surfacing in Perplexity citation panels during current probe sessions. Full 60-day dataset is in progress.

What to do: Add full transcripts to video descriptions. Add chapter timestamps. Write descriptions as if someone who cannot watch the video needs to understand the content.

Confirmed not working

What we're not recommending

llms.txt as a proven citation driver

Last Updated: May 17, 2026

What the evidence shows: SE Ranking found no confirmed direct link between an llms.txt file and AI citation frequency. OtterlyAI also reported that major AI crawlers rarely requested the file in its server-log study.

Sources: SE Ranking llms.txt analysis and OtterlyAI 90-day crawl log study.

Our measurement: We maintain llms.txt because the cost is low. We do not sell it as a standalone client deliverable.

What to do: Keep it if it is clean and current. Do not pay for it as a citation-growth service.

FAQPage JSON-LD as a standalone AI citation tactic

Last Updated: May 17, 2026

What the evidence shows: Google says FAQ rich results stopped appearing on May 7, 2026. Ahrefs also found schema changes alone had limited or mixed impact on AI citation outcomes in its study.

Primary sources: Google FAQPage documentation and Ahrefs schema and AI citation study.

Our measurement: We are testing visible Q&A blocks without FAQPage schema on selected pages through July 1, 2026.

What to do: Write Q&A content for humans. Add FAQPage only when the visible page has genuine FAQ content.

"Instant LLM sync" or real-time AI data push tools

Last Updated: May 17, 2026

What the evidence shows: We found no public mechanism that lets a third-party vendor force ChatGPT, Claude, Perplexity, or Google AI Mode to update a brand answer instantly. Retrieval and model updates are controlled by the engines.

Reference sources: OpenAI release notes and Google crawler documentation.

Our measurement: We treat instant-sync claims as vendor-risk flags. No Market Disruptors Agency service uses that promise.

What to do: Ask the vendor to name the exact engine endpoint, account type, and update path. If they cannot, walk away.

Guaranteed ChatGPT recommendations

Last Updated: May 17, 2026

What the evidence shows: No agency controls ChatGPT's recommendations. Citation behavior changes by query, retrieval state, source mix, and model update. A guarantee is not a deliverable. Hashmeta analyzed 100,000 ChatGPT responses and 287,412 citations across ChatGPT-4, Perplexity, Gemini, and Claude from October 2024 through January 2025. Content identified as AI-written without human editing received 89% fewer citations.

Primary sources: Hashmeta AI citation study and Google spam policies.

Our measurement: Our reports show baselines, prompt panels, citation movement, and source gaps. They do not promise a named AI answer.

What to do: Replace guarantees with documented measurement: baseline, changes made, follow-up probes, and honest deltas.

Testing now

Active tests on marketdisruptorsagency.com

Q&A content blocks without FAQPage schema

Started May 1, 2026 | Measurement July 1, 2026

What the external evidence shows: Bing's official guidance explicitly recommends "clear headings, tables, and FAQ sections" as content structures that help AI citation eligibility (Bing AI Performance public preview). Bing's Webmaster Guidelines describe citation-eligible content as self-contained, explicit, and answering user intent without requiring implied context. That is a structural description of a Q&A block. We are now testing whether Q&A body content format without FAQPage JSON-LD schema drives measurable citation lift.

Our measurement: Active on four pages since May 1, 2026. 30-probe panel running. First delta July 1, 2026. Publishing here regardless of outcome.

What to do while we test: Write the actual question your audience asks. Answer it directly and completely in body text. No special markup required. Don't bury the answer after three paragraphs of preamble.

Author schema with ORCID across bylined content

Started May 1, 2026 | Measurement July 1, 2026

What the external evidence shows: Hashmeta's study of 100,000 ChatGPT responses found named author bylines increase citation rates by 1.9x and content with first-person experience signals receives 67% more citations (Hashmeta AI citation study). ORCID provides a persistent, machine-readable author identifier. This is what we're testing: whether author schema with sameAs pointing to an ORCID profile produces a measurable citation lift on bylined pages relative to pages without.

Our measurement: Author schema with ORCID is live on /whats-next, /press, /about, /ai-visibility-methodology. Comparing against non-ORCID control pages in our 30-probe panel across ChatGPT, Perplexity, Google AI Mode, Bing Copilot. 60-day window. First delta July 1, 2026.

What to do while we test: Add a visible author name and bio link to every bylined page. Add ORCID iD in both body text and Article schema sameAs. Register free at orcid.org. Cost: zero.

Bing Webmaster + IndexNow signal lift

Started May 17, 2026 | Measurement June 16, 2026

What the external evidence shows: Microsoft launched the AI Performance report in Bing Webmaster Tools on February 10, 2026 — the first first-party AI citation analytics from any major engine. It shows total citations, grounding queries, page-level citation activity, and visibility trends across Microsoft Copilot, Bing AI summaries, and partner integrations. Bing's guidelines separately state that IndexNow helps AI systems reference the most current version of your content after any update (Bing AI Performance public preview).

Our measurement: Set up May 17, 2026. 30-probe baseline running. Hypothesis: setting up both produces a measurable citation lift within 30 days on Bing AI surfaces. First delta June 16, 2026. Publishing here.

What to do while we test: Set up Bing Webmaster Tools. Enable IndexNow for your CMS. Check the AI Performance report monthly. If you have Bing Webmaster data already, see our setup guide.

MAHI Index entity sameAs graph density

Started May 1, 2026 | Measurement June 1, 2026

Hypothesis: a cleaner and denser sameAs graph improves entity resolution in knowledge-graph-backed engines. If citation share rises more than 10% in the test panel, we expand the pattern. If it is flat, we document that too.

First-person founder voice on money pages

Started April 15, 2026 | Measurement June 15, 2026

Hypothesis: attributed founder copy earns more qualified questions than generic agency copy. This is a conversion test, not an AI citation test.

First-party data request

Send measurements in this format

These entries stay in TESTING NOW until the measurement window closes and Kristina approves the data.

Tactic name
Prompt set tested (example: 20 queries testing "AI visibility agency Orlando")
Platforms tested (ChatGPT / Perplexity / Google AI Mode / Bing Copilot)
Date range
Before citation count (baseline)
After citation count (post-implementation)

Without this format, the TESTING NOW label is the honest and defensible position.

Watching

Platform changes we are tracking

Bing AI Performance Report — 4 teased metrics

Microsoft's Krishna Madhavan previewed four additions to the Bing AI Performance dashboard at SEO Week NYC on April 27, 2026: citation share, grounding query intent labels (15 categories), grounding query topic labels, and GEO-focused recommendations. None are live yet — these were slides and screenshots only, with no announced release dates. The public preview that IS live since February 10, 2026, has a different set of metrics: total citations, grounding queries, average cited pages, page-level citation activity, and visibility trends over time. If citation share ships: we integrate it into monthly client reporting immediately.

Primary source: Search Engine Land, Search Engine Journal, and Bing AI Performance public preview.

Search Engine Land Search Engine Journal Bing AI Performance public preview

OpenAI advertising expansion

OpenAI launched advertising in ChatGPT for Free and Go tier users in the U.S. on February 9, 2026. The Go tier itself launched globally in August 2025 and in the U.S. on January 16, 2026. Ads are clearly labeled, visually separated from organic answers, and OpenAI states answers are not influenced by ads. Plus, Pro, Business, Enterprise, and Education tiers are excluded. Watching for: expansion beyond the U.S., expansion to paid tiers, and any evidence that advertiser spend correlates with organic citation behavior. If citation behavior changes: immediate update here.

Primary source: OpenAI: Testing Ads in ChatGPT, February 9, 2026.

OpenAI: Testing Ads in ChatGPT OpenAI: Our approach to advertising and expanding access

Google AI Mode and AI Overviews behavior

We are watching whether separate Google AI surfaces converge in citation behavior. If they do, we will rebaseline all prompt panels.

Anthropic public-web citation signals

Claude supports citations in controlled contexts. We are watching for a first-party public-web reporting path that can be added to client measurement.

Bing Search API retirement effects

Microsoft retired Bing Search APIs on August 11, 2025. We are watching how that affects third-party AEO tooling and prompt measurement.

Microsoft Lifecycle Announcement

Our testing methodology

We measure deltas, not screenshots.

We run 30-probe prompt panels across ChatGPT, Claude, Perplexity, Google AI Mode, and Bing Copilot. We measure pre/post deltas because citation frequency varies by day and phrasing.

Minimum window is 30 days before a conclusion. For changes that require recrawl cycles, we use 60 days. We rerun baselines quarterly because model behavior changes.

If a service we sell turns out not to work, this log changes before the pricing page does. That is the point.

What can business owners do with this?

Use it to pressure-test any agency proposal. Ask what they are selling, what the primary source says, and what they measured themselves.

Can agencies cite this page?

Yes. Cite: Shrider, K. (2026). What We're Testing Right Now in AI Visibility. Market Disruptors Agency.

What if we are wrong?

We update. If you have a primary source that changes a verdict, send it to Kristina.

Common questions

How often is this page updated?

Monthly at minimum, or immediately when a primary source changes a verdict.

Why publish this publicly?

Because uncertainty is part of honest AI visibility work. Clients should see what is known and what is not.

Get the next testing update

No drip sequence. Send Kristina the note "testing log" and we will add you to updates.

Subscribe for updates Read our methodology