Voice search and competitive analysis measures how often a brand's pages are chosen as the spoken answer, using query sets, share of voice, and local search visibility rather than raw traffic totals.
Spoken queries change the unit of competition. A typed search returns ten blue links and a map pack; a spoken query usually returns one answer, read aloud, with no second result to fall back on. That single-answer format is why competitive benchmarking for voice has to count wins and losses per question rather than per keyword.
Three things make the measurement problem harder than a standard rank report. First, the assistant may answer from a featured snippet, a Google Business Profile entry, or a knowledge panel, so the winning surface is not always a web page. Second, the query itself is conversational, so the same intent arrives in many phrasings. Third, the result is often invisible in analytics because the assistant reads the answer without a click.
What spoken queries change about competitive benchmarking
A rank tracker reports position. A voice benchmark reports selection: which source the assistant chose, and whether that source was the brand or a competitor. Those are different questions, and a page can rank third yet still be the cited answer if it holds the clearest short passage.
Conversational phrasing also widens the query set. A typed search might be "dry cleaning Kuching"; a spoken one is closer to "where can I get dry cleaning near me in Kuching". Both point at the same service, but the second carries more words, more local intent, and a stronger pull toward map and profile data.
That shift favours businesses with clean local signals. In the Sinar Saredah Sdn Bhd case study, Blackstone Intelligence optimised Google Business Profiles and the website for hyper-local, intent-driven keywords, built location-specific landing pages, added schema markup, and ran review generation campaigns. Local search visibility increased by 420%, and the client reached the #1 spot in the Google Local Pack for their primary locations. The case study reports local search visibility and Local Pack position; it does not claim those results came from spoken queries.
How share of voice is calculated for spoken results
Share of voice is the share of a defined query set that a brand wins. BrightEdge describes it as a way to see which competitor holds the greatest share of voice in organic search, and its product pages frame the metric around competitive position, click behaviour, and week-over-week gains and losses. That structure transfers to voice work with one adjustment: the win condition becomes "was this source selected as the answer" rather than "did this page rank".
The calculation itself is simple arithmetic. Divide the number of queries in the set where the brand was the selected source by the total number of queries in the set, then express the result as a percentage. The difficulty sits in the query set and the observation method, not the division.
Three constraints shape any honest version of this number. The query set must be fixed and dated, because a changing set makes period-on-period comparison meaningless. The observation method must be stated, because a manual check on one device in one location is not the same evidence as a logged impression count. And the assistant must be named, because Google, Siri, and Alexa draw on different surfaces and will not agree on the same query.
Featured snippets and schema markup matter here because they give an assistant a clean, self-contained passage to read. A page that answers a question in one short paragraph, marked up so the answer is machine-readable, is easier to select than a page that buries the same answer in the middle of a long section.
Building a voice search and competitive analysis baseline in Malaysia
A baseline is only useful if it can be repeated. The sequence below produces a dated, comparable record rather than a one-off snapshot.
- Write down the question set. Use 20 to 50 conversational questions that real customers would ask aloud, grouped by service and by location.
- Name the assistants and surfaces being checked, and record the device and location used for each check.
- For every question, record which source was selected as the answer and whether it belonged to the brand, a named competitor, or a directory.
- Record the surface the answer came from, such as a featured snippet, a Google Business Profile entry, or a map result.
- Calculate share of voice by dividing questions won by total questions in the set.
- Repeat the same set on the same surfaces at a fixed interval, and log the date of each run.
- Compare gains and losses question by question, then trace each loss back to the page or profile that should have answered it.
Malaysian businesses face a specific constraint here: there is no supplied source giving Malaysian voice search query volumes, device mix, or assistant usage rates, and no supplied benchmark for what a normal share of voice looks like in this market. That absence changes the method rather than blocking it. Without an external benchmark, the useful comparison is the brand against its own named competitors on a fixed question set, measured over time. A first-party baseline beats an imported statistic that does not describe the local market.
Local signals carry unusual weight in that baseline. Google Business Profile accuracy, consistent name, address, and phone details, location-specific pages, and review volume all feed the surfaces an assistant reads for "near me" questions. The Sinar Saredah work shows what that combination looks like in practice: profile optimisation, location pages, schema markup, and review generation, with the Local Pack result as the reported outcome.
What Blackstone Intelligence has measured on local search visibility
Blackstone Intelligence is a Kuching-based AI systems and digital growth agency operated by Blackstone Consultancy Sdn Bhd, founded by Anton Dandot. Its public profile describes SEO and search systems covering local search optimisation, service-page structuring, search-ready content systems, and Google ranking support.
The measurable local search record comes from named case studies rather than a voice-specific dataset. For Sinar Saredah Sdn Bhd, a commercial and residential laundry and dry cleaning service, the client was buried on page 3 or 4 of Google results for searches like "dry cleaning near me". After the work described above, local search visibility increased by 420%, the client reached the #1 spot in the Google Local Pack for their primary locations, and B2B contracts grew by 85%, including long-term agreements with boutique hotels and restaurant chains. The same engagement reported a 3.5x return on ad spend on social advertising and a 65% reduction in cost per acquisition.
For Eyonic Sdn Bhd, the profile records local SEO work for CCTV, access control, and security services, with page one reached for targeted local search terms within 20 days. Neither case study reports spoken-query impressions, assistant referrals, or a voice share of voice figure, and no supplied source confirms whether Blackstone Intelligence offers a distinct voice search service or how such work would be scoped.
What the record does support is the underlying capability that voice visibility depends on: profile optimisation, location pages, schema markup, review generation, and search-ready page structure. Those are the inputs an assistant reads. Whether they convert into spoken-answer selection is a separate question that needs its own measurement.
Limits of competitor pages as evidence for voice search claims
Competitor pages are useful for structure and topic coverage. They are not proof of performance. Across the seven analysed competitor pages, none used the complete exact-match query in its H1 or body, the median length was 3,030 words with a median of 17 headings, and five of seven carried lists and FAQ blocks. That tells a writer what the topic usually covers. It says nothing about whether any of those pages actually wins spoken answers.
Several specific claims in that set cannot be verified from the supplied evidence. The ALM Corp 2026 guide and the BrightEdge share of voice pages make performance and measurement claims that no supplied source confirms. BrightEdge's own material describes share of voice as a competitive reporting metric for organic search; it does not establish a voice search benchmark, and it should not be cited as one.
Three pages in the competitor set returned errors during analysis, which is a reminder that a competitor page can be cited long after it stops resolving. A claim copied from a dead page is still unverified.
The practical rule is to separate two kinds of statement. "This competitor covers featured snippets and schema markup" is an observation about content. "This competitor wins voice search in Malaysia" is a performance claim that needs first-party measurement, a named method, and a date range. Only the first kind belongs in a competitive analysis built on supplied evidence.
Voice search and competitive analysis come together at the point where a brand can state, with a dated question set and a named method, which questions it wins, which competitors win the rest, and which page or profile should have answered. That record is defensible. A borrowed statistic is not.