Method: State of Play, US iGaming visibility, edition one

What this page is for

This page explains how TBD.Media measured the results in State of Play, the study it publishes. Use the method and supporting evidence to examine or challenge a published figure.

The method version is v2.1. The fixed search list version is 1.0. The study’s brand record versions are nj-v3.1 for New Jersey and mi-v2.1 for Michigan. The record of published figures was built on 18 September 2026.

The searches and how we chose them

The fixed search list contains 211 distinct searches across New Jersey and Michigan. Each search belongs to one table.

Scroll sideways to see every column. The first column stays in place.

Rows show distinct search counts by state and table

These counts describe distinct searches. Checking the same search on another date creates another observation, but does not add a distinct search. The list does not cover every search customers make.

How we chose the searches

TBD.Media gathered candidate searches from DataForSEO and its export from Ahrefs. DataForSEO is a search research service used here to find candidate searches and obtain monthly search volumes for each location. Ahrefs is a search research service whose export supplied additional candidate searches. The sources included searches associated with licensed brands and suggestions based on starting search phrases. TBD.Media removed duplicates before applying the inclusion rules.

We used these licensed-brand websites to build the starting pool: fanduel.com, draftkings.com, betmgm.com, hardrock.bet, borgataonline.com, caesars.com, resortscasino.com, sportsbook.fanduel.com, sportsbook.draftkings.com, nj.betmgm.com, mi.betmgm.com, nj.betrivers.com, mi.betrivers.com, playnj.com, playmichigan.com, nj.bet365.com and mi.bet365.com. We used separate starting phrases for casino searches (nj online casino, new jersey online casino, nj casino bonus, nj online slots, nj live dealer casino, michigan online casino, mi online casino, michigan casino bonus, michigan online slots and michigan live dealer casino) and sportsbook searches (new jersey sports betting, nj sportsbook, nj sportsbook promo, nj betting app, nj parlay, michigan sports betting, michigan sportsbook, michigan sportsbook promo, michigan betting app and michigan parlay).

Included searches must show state intent in their wording or results and meet the method’s minimum monthly demand for that location. The core groups exclude searches containing brand names.

Scroll sideways to see every column. The first column stays in place.

Casino and sportsbook searches cover these topic groups

Team and event searches have separate tables because they cover different topics. Their scores do not contribute to the main sportsbook ranking.

How monthly demand affects search weight

A search weight determines how much a search contributes to the Visibility score. Higher monthly demand gives a search more influence. DataForSEO supplies the monthly search volume for the relevant location.

A higher weight makes a search count for more in the score, in proportion to how many people search for it.

Scroll sideways to see every column. The first column stays in place.

Monthly search ranges determine the assigned weights

The full search-list appendix gives each search’s text, state, product, table, monthly volume and assigned weight.

Refreshing the list

The method commits to replacing part of the search list for each edition, using the same selection process. Comparisons between editions apply at market level.

The collection rounds and their dates

Scroll sideways to see every column. The first column stays in place.

Collection and review dates show the coverage and purpose of each check

We collected mobile Google results with the location set to New Jersey or Michigan. We recorded the first 20 unpaid Google results, excluding advertisements, with Google’s AI-generated summaries, called AI Overviews, enabled.

A search entered the third collection pass when a domain belonging to a brand in that search’s state register moved by more than five places between the first two passes.

Only 37 of 211 searches received a third collection. See the selected-search list for the searches covered. Not every search received three checks.

The Visibility score uses the first two full collections and the selected third collection. The separate casino AI Overview figures use 15 September 2026. The sportsbook AI Overview figures use 10 September 2026 and 13 September 2026.

The human reviews check selected searches. They are not additional full collections.

Scoring in plain words

The Visibility score measures search presence on a scale from zero to one hundred. It compares points earned with the maximum weighted points available. It is not a percentage of searches, traffic or customers.

An AI Overview is Google’s generated summary within search results. AI Overview citation means its source list includes a website address recorded for the brand. AI Overview mention means Google’s own answer text names the brand without a source-list citation for that brand.

An “AI Overview citation” earns four points. An “AI Overview mention” earns two points. If both occur for the same brand and search, only the “AI Overview citation” earns points.

How we calculate the Visibility score

1. Give each recorded website address points for its median position among unpaid results. The median is the middle recorded position after sorting; with two positions, use their average.
2. Reduce those points evenly from ten points at the first position to one point at the twentieth position.
3. Add four points when the brand receives an “AI Overview citation”.
4. Give two points when the brand receives an “AI Overview mention”.
5. Where both occur, award only the points for the “AI Overview citation”.
6. Add one point for a matching website address in an app listing box or Google’s Top stories.
7. Cap the brand’s total at ten points per search.
8. Multiply each search’s capped points by its search-volume weight.
9. Divide the total by the maximum weighted points available across all searches in that table.
10. Multiply the result by one hundred.

For the unpaid-result calculation, the first and twentieth positions define the position range. The highest and lowest point values are ten points and one point.

Unpaid-result points = highest points − ((median position − first position) × (highest points − lowest points) ÷ (last position − first position)).

A brand earns ten points at the first position and one point at the twentieth, with points falling evenly between those positions.

A website address with no recorded position among the first twenty unpaid results earns no unpaid-result points.

Visibility score = one hundred × total weighted points earned ÷ maximum weighted points available.

The maximum uses ten points multiplied by each search’s weight, summed across every search in the table.

Worked example, not a finding

Suppose a table contains two searches. The brand earns eight points on a search weighted two, and four points on a search weighted one.

Weighted points = (eight × two) + (four × one) = twenty

Maximum weighted points = (ten × two) + (ten × one) = thirty

Visibility score = one hundred × twenty ÷ thirty = sixty-six point seven out of one hundred, after rounding

Rules that affect the calculation

1. Unpaid-result points add per website address. A domain is a website address recorded for the brand. Each recorded domain can contribute points before the brand’s per-search cap of ten points.
2. AI Overview mention uses Google’s own answer. Edition one matches the brand name or a permitted alternative name in Google’s own answer headings and paragraphs. Titles, short extracts, source names and addresses of cited third-party pages do not earn points for an “AI Overview mention”. A link within the answer without a source-list citation counts as an “AI Overview mention” when it meets this rule. The rule for “AI Overview citation” is unchanged.
3. Absent rounds supply no position. The median uses only rounds where the domain was recorded. This is a known limit of edition one and can favour a domain that appears intermittently. The third collection covered 37 of 211 searches. Edition two will use full passes across all collection rounds.
4. Name matching uses the relevant state and product record. We use the official name and permitted alternative names from that brand’s corresponding record. We do not borrow alternative names from other rows.
5. App listings and Top stories require a domain match. A brand name alone in either feature earns no points.
6. One recorded unpaid-result appearance means one domain, one search and one collection round. Repeating a search on another date can add an appearance, but it remains one distinct search.
7. The study’s state record determines each site’s type. An operator may hold a licence in one state but not the other. Review and comparison sites run by media groups, Affiliate sites and News media are separate classes. Each site’s appearances belong to one class.
8. Alternative brand names must contain at least two words. This reduces accidental matches. The restriction does not apply to official brand names.

When one access method is blocked

The fact-page readiness check tries several routes to read a page. A firewall can block one route while another returns the real page.

The collection label “blocked by policy” counts as detected for the automated-access check. The label means a plain automated visit was refused, then a more patient visit behaving like an ordinary browser read the real page. A refusal with no successful visit at all earns no credit.

This does not establish that every automated system can reach the page. The method records this access rule as agreed on the eleventh of September, two thousand and twenty-six.

Bands and the margin

The Visibility score measures a brand’s presence in the measured Google results on a scale from zero to one hundred.

Scroll sideways to see every column. The first column stays in place.

Visibility bands group scores by measured search presence

Near invisible describes the searches in the relevant table. It does not mean customers cannot find the brand elsewhere.

Fact-page readiness summarises seven checks of access to a brand’s information and the information the checks find. These cover automated access, the state landing page, licence disclosure, responsible gambling, organisation data, welcome offer terms and operator identity.

The weights total twenty-five raw points. We multiply the raw total by one hundred and divide by twenty-five to assign a readiness band.

Scroll sideways to see every column. The first column stays in place.

Fact-page readiness bands group converted scores

A brand needs measurable automated access and at least five of the seven checks measurable before it receives a readiness band. Pending means supporting evidence awaits human confirmation. Not rated means the measurement requirements are not met.

We publish fact-page readiness bands for New Jersey only. Readiness does not establish legal or regulatory compliance.

What the margin means

The practical margin is plus or minus 2 Visibility points. It reflects observed variation between collection rounds, including the selected third collection.

The third collection covered 37 of 211 searches. A search entered the third collection pass when a domain belonging to a brand in that search’s state register moved by more than five places between the first two passes.

The margin is not a formal statistical confidence interval. Small score differences do not establish a meaningful lead.

Where two scores differ by less than the stated margin of plus or minus 2 Visibility points, treat their order as not firm. The tables contain such pairs.

The tables order ranked brands by Visibility score, highest first. Equal scores appear in alphabetical order. The tables do not merge nearby scores into tied ranks or use AI Overview citations to decide their order.

Brands scoring below 5 appear alphabetically without rank numbers.

The readiness checks and their weights

Fact-page readiness assesses whether search engines and AI systems can access and understand a brand’s key information. The seven checks contribute up to 25 raw points. Readiness does not predict a Google ranking, guarantee an AI Overview citation or establish regulatory compliance.

The pages in scope are the state landing page, about or company page, licence or legal page, responsible-gambling page and one article.

Scroll sideways to see every column. The first column stays in place.

Each readiness check lists its weight, required evidence and human review

The table describes the review requirements. It does not claim that a person reviewed every check for every brand.

A detected check earns its listed weight. A measurable check with a confirmed “not detected” result earns no points. “Not measurable” is a separate status, not a confirmed absence of information.

We multiply the raw total by one hundred and divide by 25 to convert the 25 raw points to the readiness scale.

• Ready: at least 70.
• Partial: at least 40 and below 70.
• Gaps: below 40.

Worked example, not a finding: Suppose automated access, state landing page, licence disclosure, responsible gambling and organisation data are detected.

Welcome offer terms and operator identity have confirmed “not detected” results. The brand scores 4 + 4 + 4 + 4 + 3 + 0 + 0 = 19 raw points.

Readiness score = 19 × 100 ÷ 25 = 76 out of 100

The brand receives Ready because 76 is at least 70. This example assumes all seven checks are measurable and the required confirmation is complete.

A brand receives a band only when automated access and at least five of the seven checks are measurable.

• Pending: Supporting evidence still needs human confirmation.
• Not rated: Automated access is not measurable, or fewer than five of the seven checks are measurable.
• Not detected: The check looked for the required evidence and did not find it.
• Not measurable: The check could not produce a reliable result.

Missing or uncertain evidence needs confirmation. It does not establish a regulatory breach.

The “blocked by policy” access rule awards detected credit only when another collection route returns the real page. Successful access through one route does not establish access for every automated system.

We publish fact-page readiness bands for New Jersey only. We publish no per-check licence or responsible-gambling result for any brand.

We publish no Michigan readiness result, status or check count for any brand. Michigan readiness remains private because its newest records have not received an independent review that challenges the findings.

The register and its sources

The register is the study’s brand record, built from state regulators’ published lists. It is separate from the regulator’s own licensing list.

New Jersey. Eligibility follows the New Jersey Division of Gaming Enforcement’s licensing records for the relevant product. Register version nj-v3.1 contains 46 rows.

Source pages: https://www.njoag.gov/about/divisions-and-offices/division-of-gaming-enforcement-home/internet-gaming-sites/ and https://www.njoag.gov/about/divisions-and-offices/division-of-gaming-enforcement-home/sports-wagering/.

Edition one includes only New Jersey product rows ready for measurement. It excludes exchange products because this is a fixed-odds sportsbook study. A sportsbook row is also outside this edition where the brand is measurable only for casino. Any product row not validated in time is excluded.

Michigan. Eligibility follows the Michigan Gaming Control Board’s authorised-provider page, fetched on the seventeenth of September, two thousand and twenty-six. Register version mi-v2.1 contains 32 rows.

Michigan launch approvals. TBD.Media’s team found the regulator pages supporting the inclusion of bet365 and Hard Rock Bet. The team retrieved the regulator’s releases and saved the evidence.

The regulator granted bet365’s launch approval on 17 April 2026. It granted Hard Rock Bet’s launch approval on 2 December 2025.

Both approval dates precede the study’s September 2026 collection. Approval dates describe regulatory decisions; collection dates describe when we gathered search results. TBD.Media’s team records which regulator sources it checked and when it checked them.

The Michigan tables exclude operators no longer licensed there.

What each row holds. Each row records the licensee or platform, legal operator, website addresses, sub-sites and permitted alternative brand names. Alternative names used for mention matching must contain at least two words. Rows also record addresses for the state, about, licence, responsible-gambling and article pages.

Changes to the register. The register change log records updated page addresses, licensing evidence and brand records. It distinguishes confirmed addresses from unresolved evidence and records corrections to earlier results.

The hand checks

TBD.Media’s team reviewed selected searches for 20 rows. The team selected the highest-ranked rows from the New Jersey and Michigan casino and main sportsbook tables. The team and event tables have no hand-checked rows.

The team reviewed New Jersey searches on 16 September 2026 and Michigan searches on 17 September 2026. These were human reviews, not additional full collections.

Some checks used a UK desktop browser with Google’s location set to the relevant state. Others used fresh results through a US mobile connection. The scored collection used mobile results with the state location set.

A UK desktop check and a US mobile check can show different positions. Where they disagreed, the reviewer treated the US mobile check as the closer match to the collection conditions. The row note records that judgement.

A hand check confirms only the recorded positions on the named searches on that date. It does not independently recalculate the Visibility score across every search in the table. Any comment on the Visibility band is the reviewer’s judgement, not a separate measurement of that band.

A hand check does not confirm a fact-page readiness band or another row. Read each row’s Human-check status with its note to see the searches, date and conditions covered.

Pre-notification

No notices were sent to brands before publication. For a brand’s fact pages, the study publishes a readiness band only, and only for New Jersey. The band summarises the checks on those pages. The study publishes no per-check licence result and no per-check responsible-gambling result for any brand. Individual results remain in the private data.

Three New Jersey brands carry a licence check of “not detected” in the private data. These brands are not named on any public page. In each case, the page served to a visitor outside New Jersey showed another state’s wording. “Not detected” means the check did not find the required evidence on the page it received. It does not establish that the brand lacks a New Jersey licence.

The earlier method text promised notice to a brand carrying such a result. The owner decided that notice was not required because the study publishes no per-check result for any brand. This decision concerns what the study publishes; it does not change the licence checks recorded in the private data.

The public readiness band summarises the fact-page assessment, without identifying the result of an individual check. Any brand may ask for its own row and the evidence behind it through the corrections contact: [email protected].

Requesting a correction

To request a correction, contact [email protected]. Identify the statement you believe is wrong and provide evidence that supports your correction.

Versions, decisions and known limits

Scroll sideways to see every column. The first column stays in place.

Study files and brand records have these recorded versions

TBD.Media compiled the study’s published figures on 18 September 2026.

Adopted AI Overview mention rule

An AI Overview is Google’s AI-generated summary within search results. Edition one counts an AI Overview mention only when Google’s own answer text contains the brand’s official name or a permitted alternative name. This includes answer headings and paragraphs. Titles, short extracts, source names and addresses of cited third-party pages do not qualify.

TBD.Media applied this rule to every collection round in edition one and kept no result collected under the earlier rule.

An AI Overview citation means the source list includes a website address recorded for the brand. The AI Overview citation rule is unchanged. This decision is settled for edition one.

Known limit: absent collection rounds

The median position uses only rounds where the domain was recorded. The calculation ignores a round where the domain is absent instead of scoring it as a no-show. This can favour a domain that appears intermittently.

The owner retained this rule for edition one. The third collection covered 37 of 211 searches. A search entered the third collection pass when a domain belonging to a brand in that search’s state register moved by more than five places between the first two passes.

Edition two will use full passes across all collection rounds. The treatment of absent rounds is a stated limit of edition one, not an unresolved decision for this edition.

Appendix: the full search list

The fixed list contains 211 searches, with monthly demand for each location and the weight each search receives. We repeated 37 in the third collection.

New Jersey, online casino

Scroll sideways to see every column. The first column stays in place.

New Jersey, online sportsbook

Scroll sideways to see every column. The first column stays in place.

New Jersey, team and event searches

Scroll sideways to see every column. The first column stays in place.

Michigan, online casino

Scroll sideways to see every column. The first column stays in place.

Michigan, online sportsbook

Scroll sideways to see every column. The first column stays in place.

Michigan, team and event searches

Scroll sideways to see every column. The first column stays in place.

Appendix: the searches collected a third time

We repeated 37 of 211 searches in the third collection. On each, a licensed brand’s website moved more than five places between the earlier collections.

Scroll sideways to see every column. The first column stays in place.