How OBScoreboard went from zero to 129 AI answers citing its pages

The situation

OBScoreboard sells live scoreboard overlays that plug into OBS, vMix and Streamlabs. The buyers are sports streamers, school and club media teams, small leagues and production crews who need a clean score bug on a live stream without hiring a graphics operator. It is a classic long-tail SaaS: thousands of tiny, very specific searches ('cricket scoreboard for OBS', 'vMix basketball scorebug', 'how to add a live score to Streamlabs') and almost no single head term worth fighting for.

That long tail is exactly where AI answers now sit. When someone asks Google or ChatGPT how to put a live score on a stream, the engine writes a step-by-step answer and cites a handful of sources. Before this program, obscoreboard.com was cited in zero AI responses. The site had a Domain Rating of 5 and 145 referring domains, which in practice meant Google had little reason to trust it over a forum thread or a YouTube tutorial from someone else.

The goal was not traffic for its own sake. It was to become the source AI engines quote when a streamer asks how to do the exact job OBScoreboard does, so the product name appears inside the answer itself.

01How does Google AI Overviews actually choose which pages to cite?

Google does not run one search when it builds an AI Overview. It runs a query fan-out: the model breaks the question into several sub-queries ('what software overlays scores on OBS', 'is there a free option', 'how to connect a scoreboard to vMix', 'does it work on mobile streams'), retrieves results for each, and then stitches an answer from the passages that best answer each sub-question. AI Mode does the same thing with even more sub-queries.

Two consequences follow. First, a page can be cited for a question it never ranked on page one for, because it won a sub-query. Second, retrieval is passage-level. Google is not quoting your page, it is quoting a short block of a few sentences on your page that answers one sub-question cleanly. If that answer is buried under a hero image, a sales intro and three paragraphs of context, a competitor's tidier passage wins.

The third filter is trust. Among the passages that answer the sub-query, Google leans toward sources from its normal ranking system, which still weighs links, entity clarity and site reputation. That is why a technically perfect page on a DR 5 site loses to a mediocre Reddit thread: Reddit has the trust, the small SaaS does not yet.

• Answer the sub-question in the first two sentences under each heading, then elaborate.
• Write headings as the literal question a user would ask ('How do I add a live score to OBS?').
• Use numbered steps for procedures. Engines lift ordered lists almost intact.
• Mark up the page with SoftwareApplication, HowTo-style structure and FAQPage where it matches visible content, so the entity is unambiguous.

02Why build a page for every streaming tool and every sport?

A SaaS like OBScoreboard has a natural matrix: three host tools (OBS, vMix, Streamlabs) multiplied by every sport a streamer covers (football, basketball, cricket, volleyball, hockey, esports and more). Each cell of that matrix is a distinct question with distinct intent, and each one is a fan-out sub-query waiting for a clean answer.

The play is a programmatic page set built on a real template, not a spun one. Each page covers one tool and sport combination with the specifics that actually differ: how the browser source or input is added in that tool, which fields that sport needs (overs and wickets, quarters and fouls, sets and serve), what the overlay looks like, and the common failure for that setup. The shared parts come from a component, the differing parts are written by someone who has actually set the overlay up.

This is where most programmatic SEO fails. Pages that differ only by a swapped keyword are near-duplicates, Google folds them together, and AI engines have nothing unique to quote. Pages that each carry one genuinely different procedure become citable. The test we apply is simple: if you deleted the sport name, could you tell which page you were on? If not, the page is not ready.

• Integration hub pages: one per host tool, linking down to every sport page for that tool.
• Sport hub pages: one per sport, linking across to every tool page for that sport.
• Troubleshooting passages per tool ('overlay not updating in vMix', 'browser source shows black in OBS').
• A comparison passage on each hub that names alternatives honestly, because 'what is the best' is always one of the fan-out sub-queries.

03Why do referring domains correlate with AI Overview citations?

Over the same period the AI citations appeared, OBScoreboard's referring domains went from 145 to 921 and Domain Rating from 5 to 12. That is not a coincidence. AI Overviews draw from pages Google already considers rankable, and the count of distinct, relevant sites linking to you is still one of the strongest inputs into that judgement.

The key word is distinct. Ten thousand links from one site add little. Hundreds of different domains each linking once, from streaming blogs, tool directories, league sites, school media pages and tutorial creators, is what moves the needle, because each one is an independent vote that the site belongs in the conversation about live sports streaming.

There is a second, quieter effect. Many of those linking pages mention OBScoreboard by name next to words like 'OBS', 'vMix', 'scoreboard' and 'overlay'. Those co-occurrences teach both Google and the language models behind ChatGPT and Gemini what the brand is. Entity understanding is built from mentions as much as from links.

• Directory and marketplace listings: streaming tool directories, SaaS directories, and alternative-to pages.
• Integration ecosystem: plugin lists, community resource pages and 'tools for streamers' roundups.
• Customer sites: leagues, clubs and schools that stream their games and credit the overlay they use.
• Creator links: tutorial makers who link the tool in their video descriptions and blog write-ups.

05Why do AI engines quote Reddit threads so often?

Reddit threads have three properties AI engines love. They are phrased as real questions in the user's own words, which match fan-out sub-queries closely. They contain multiple independent answers, which reads as consensus. And the best answers are short, specific and experiential: 'I use X with OBS, set the browser source to 1920x1080 and it updates every second.' That is a perfectly sized, quotable passage.

Google's licensing deal with Reddit and the platform's general visibility in search mean those threads are also crawled fast and ranked well. For a small SaaS, the practical lesson is that the Reddit thread answering 'best scoreboard overlay for OBS' is part of your site's AI footprint whether you like it or not. Being present and accurate in that thread, with disclosure, is often a faster route into AI answers than anything on your own domain.

06How do YouTube tutorials get a SaaS into AI answers?

Setup questions in streaming are visual, and YouTube owns them. Google surfaces videos inside AI Overviews and AI Mode, and the engines read the transcript, the title, the description and the chapter markers. A tutorial titled with the exact question and chaptered by step gives the engine timestamped passages to cite.

The play is one short tutorial per high-intent cell of the matrix: 'Add a live basketball scoreboard in OBS in 3 minutes', 'vMix scoreboard input setup', 'Streamlabs score overlay for school sports'. Say the steps out loud in plain words, because the transcript is the text the engine reads. Add chapters named as sub-questions. Put the matching page link first in the description, and embed the video on that page so the page and the video reinforce each other.

Creator partnerships multiply this. A streaming-gear YouTuber who shows the overlay in a 'my stream setup' video creates a third-party mention, a description link and a transcript passage in one go.

• Title = the question. Chapters = the steps. Description first line = the answer in one sentence.
• Upload accurate captions rather than relying on auto-captions that mangle product names.
• Shorts showing the overlay live are discovery, long tutorials are what gets cited.